inode.c 231.4 KB
Newer Older
C
Chris Mason 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * Copyright (C) 2007 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 */

19
#include <linux/kernel.h>
20
#include <linux/bio.h>
C
Chris Mason 已提交
21
#include <linux/buffer_head.h>
S
Sage Weil 已提交
22
#include <linux/file.h>
C
Chris Mason 已提交
23 24 25 26 27 28 29 30 31 32 33 34
#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/highmem.h>
#include <linux/time.h>
#include <linux/init.h>
#include <linux/string.h>
#include <linux/backing-dev.h>
#include <linux/mpage.h>
#include <linux/swap.h>
#include <linux/writeback.h>
#include <linux/statfs.h>
#include <linux/compat.h>
C
Chris Mason 已提交
35
#include <linux/bit_spinlock.h>
J
Josef Bacik 已提交
36
#include <linux/xattr.h>
J
Josef Bacik 已提交
37
#include <linux/posix_acl.h>
Y
Yan Zheng 已提交
38
#include <linux/falloc.h>
39
#include <linux/slab.h>
40
#include <linux/ratelimit.h>
41
#include <linux/mount.h>
42
#include <linux/btrfs.h>
D
David Woodhouse 已提交
43
#include <linux/blkdev.h>
C
Chris Mason 已提交
44
#include "compat.h"
C
Chris Mason 已提交
45 46 47 48 49
#include "ctree.h"
#include "disk-io.h"
#include "transaction.h"
#include "btrfs_inode.h"
#include "print-tree.h"
50
#include "ordered-data.h"
51
#include "xattr.h"
52
#include "tree-log.h"
53
#include "volumes.h"
C
Chris Mason 已提交
54
#include "compression.h"
55
#include "locking.h"
56
#include "free-space-cache.h"
57
#include "inode-map.h"
L
Liu Bo 已提交
58
#include "backref.h"
C
Chris Mason 已提交
59 60 61 62 63 64

struct btrfs_iget_args {
	u64 ino;
	struct btrfs_root *root;
};

65 66 67 68 69
static const struct inode_operations btrfs_dir_inode_operations;
static const struct inode_operations btrfs_symlink_inode_operations;
static const struct inode_operations btrfs_dir_ro_inode_operations;
static const struct inode_operations btrfs_special_inode_operations;
static const struct inode_operations btrfs_file_inode_operations;
70 71
static const struct address_space_operations btrfs_aops;
static const struct address_space_operations btrfs_symlink_aops;
72
static const struct file_operations btrfs_dir_file_operations;
73
static struct extent_io_ops btrfs_extent_io_ops;
C
Chris Mason 已提交
74 75

static struct kmem_cache *btrfs_inode_cachep;
76
static struct kmem_cache *btrfs_delalloc_work_cachep;
C
Chris Mason 已提交
77 78 79
struct kmem_cache *btrfs_trans_handle_cachep;
struct kmem_cache *btrfs_transaction_cachep;
struct kmem_cache *btrfs_path_cachep;
80
struct kmem_cache *btrfs_free_space_cachep;
C
Chris Mason 已提交
81 82 83 84 85 86 87 88 89 90 91 92

#define S_SHIFT 12
static unsigned char btrfs_type_by_mode[S_IFMT >> S_SHIFT] = {
	[S_IFREG >> S_SHIFT]	= BTRFS_FT_REG_FILE,
	[S_IFDIR >> S_SHIFT]	= BTRFS_FT_DIR,
	[S_IFCHR >> S_SHIFT]	= BTRFS_FT_CHRDEV,
	[S_IFBLK >> S_SHIFT]	= BTRFS_FT_BLKDEV,
	[S_IFIFO >> S_SHIFT]	= BTRFS_FT_FIFO,
	[S_IFSOCK >> S_SHIFT]	= BTRFS_FT_SOCK,
	[S_IFLNK >> S_SHIFT]	= BTRFS_FT_SYMLINK,
};

93
static int btrfs_setsize(struct inode *inode, struct iattr *attr);
94
static int btrfs_truncate(struct inode *inode);
95
static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent);
96 97 98 99
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written, int unlock);
100 101 102
static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
					   u64 len, u64 orig_start,
					   u64 block_start, u64 block_len,
J
Josef Bacik 已提交
103 104
					   u64 orig_block_len, u64 ram_bytes,
					   int type);
105

106 107
static int btrfs_dirty_inode(struct inode *inode);

108
static int btrfs_init_inode_security(struct btrfs_trans_handle *trans,
109 110
				     struct inode *inode,  struct inode *dir,
				     const struct qstr *qstr)
J
Jim Owens 已提交
111 112 113
{
	int err;

114
	err = btrfs_init_acl(trans, inode, dir);
J
Jim Owens 已提交
115
	if (!err)
116
		err = btrfs_xattr_security_init(trans, inode, dir, qstr);
J
Jim Owens 已提交
117 118 119
	return err;
}

C
Chris Mason 已提交
120 121 122 123 124
/*
 * this does all the hard work for inserting an inline extent into
 * the btree.  The caller should have done a btrfs_drop_extents so that
 * no overlapping inline items exist in the btree
 */
C
Chris Mason 已提交
125
static noinline int insert_inline_extent(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
126 127
				struct btrfs_root *root, struct inode *inode,
				u64 start, size_t size, size_t compressed_size,
128
				int compress_type,
C
Chris Mason 已提交
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143
				struct page **compressed_pages)
{
	struct btrfs_key key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct page *page = NULL;
	char *kaddr;
	unsigned long ptr;
	struct btrfs_file_extent_item *ei;
	int err = 0;
	int ret;
	size_t cur_size = size;
	size_t datasize;
	unsigned long offset;

144
	if (compressed_size && compressed_pages)
C
Chris Mason 已提交
145 146
		cur_size = compressed_size;

C
Chris Mason 已提交
147 148
	path = btrfs_alloc_path();
	if (!path)
C
Chris Mason 已提交
149 150
		return -ENOMEM;

151
	path->leave_spinning = 1;
C
Chris Mason 已提交
152

L
Li Zefan 已提交
153
	key.objectid = btrfs_ino(inode);
C
Chris Mason 已提交
154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174
	key.offset = start;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(cur_size);

	inode_add_bytes(inode, size);
	ret = btrfs_insert_empty_item(trans, root, path, &key,
				      datasize);
	if (ret) {
		err = ret;
		goto fail;
	}
	leaf = path->nodes[0];
	ei = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
	btrfs_set_file_extent_type(leaf, ei, BTRFS_FILE_EXTENT_INLINE);
	btrfs_set_file_extent_encryption(leaf, ei, 0);
	btrfs_set_file_extent_other_encoding(leaf, ei, 0);
	btrfs_set_file_extent_ram_bytes(leaf, ei, size);
	ptr = btrfs_file_extent_inline_start(ei);

175
	if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
176 177
		struct page *cpage;
		int i = 0;
C
Chris Mason 已提交
178
		while (compressed_size > 0) {
C
Chris Mason 已提交
179
			cpage = compressed_pages[i];
180
			cur_size = min_t(unsigned long, compressed_size,
C
Chris Mason 已提交
181 182
				       PAGE_CACHE_SIZE);

183
			kaddr = kmap_atomic(cpage);
C
Chris Mason 已提交
184
			write_extent_buffer(leaf, kaddr, ptr, cur_size);
185
			kunmap_atomic(kaddr);
C
Chris Mason 已提交
186 187 188 189 190 191

			i++;
			ptr += cur_size;
			compressed_size -= cur_size;
		}
		btrfs_set_file_extent_compression(leaf, ei,
192
						  compress_type);
C
Chris Mason 已提交
193 194 195 196
	} else {
		page = find_get_page(inode->i_mapping,
				     start >> PAGE_CACHE_SHIFT);
		btrfs_set_file_extent_compression(leaf, ei, 0);
197
		kaddr = kmap_atomic(page);
C
Chris Mason 已提交
198 199
		offset = start & (PAGE_CACHE_SIZE - 1);
		write_extent_buffer(leaf, kaddr + offset, ptr, size);
200
		kunmap_atomic(kaddr);
C
Chris Mason 已提交
201 202 203 204 205
		page_cache_release(page);
	}
	btrfs_mark_buffer_dirty(leaf);
	btrfs_free_path(path);

206 207 208 209 210 211 212 213 214
	/*
	 * we're an inline extent, so nobody can
	 * extend the file past i_size without locking
	 * a page we already have locked.
	 *
	 * We must do any isize and inode updates
	 * before we unlock the pages.  Otherwise we
	 * could end up racing with unlink.
	 */
C
Chris Mason 已提交
215
	BTRFS_I(inode)->disk_i_size = inode->i_size;
216
	ret = btrfs_update_inode(trans, root, inode);
217

218
	return ret;
C
Chris Mason 已提交
219 220 221 222 223 224 225 226 227 228 229
fail:
	btrfs_free_path(path);
	return err;
}


/*
 * conditionally insert an inline extent into the file.  This
 * does the checks required to make sure the data is small enough
 * to fit as an inline extent.
 */
230
static noinline int cow_file_range_inline(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
231 232
				 struct btrfs_root *root,
				 struct inode *inode, u64 start, u64 end,
233
				 size_t compressed_size, int compress_type,
C
Chris Mason 已提交
234 235 236 237 238
				 struct page **compressed_pages)
{
	u64 isize = i_size_read(inode);
	u64 actual_end = min(end + 1, isize);
	u64 inline_len = actual_end - start;
239
	u64 aligned_end = ALIGN(end, root->sectorsize);
C
Chris Mason 已提交
240 241 242 243 244 245 246
	u64 data_len = inline_len;
	int ret;

	if (compressed_size)
		data_len = compressed_size;

	if (start > 0 ||
C
Chris Mason 已提交
247
	    actual_end >= PAGE_CACHE_SIZE ||
C
Chris Mason 已提交
248 249 250 251 252 253 254 255
	    data_len >= BTRFS_MAX_INLINE_DATA_SIZE(root) ||
	    (!compressed_size &&
	    (actual_end & (root->sectorsize - 1)) == 0) ||
	    end + 1 < isize ||
	    data_len > root->fs_info->max_inline) {
		return 1;
	}

256
	ret = btrfs_drop_extents(trans, root, inode, start, aligned_end, 1);
257 258
	if (ret)
		return ret;
C
Chris Mason 已提交
259 260 261 262 263

	if (isize > actual_end)
		inline_len = min_t(u64, isize, actual_end);
	ret = insert_inline_extent(trans, root, inode, start,
				   inline_len, compressed_size,
264
				   compress_type, compressed_pages);
265
	if (ret && ret != -ENOSPC) {
266 267
		btrfs_abort_transaction(trans, root, ret);
		return ret;
268 269
	} else if (ret == -ENOSPC) {
		return 1;
270
	}
271

272
	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
273
	btrfs_delalloc_release_metadata(inode, end + 1 - start);
C
Chris Mason 已提交
274
	btrfs_drop_extent_cache(inode, start, aligned_end - 1, 0);
C
Chris Mason 已提交
275 276 277
	return 0;
}

278 279 280 281 282 283
struct async_extent {
	u64 start;
	u64 ram_size;
	u64 compressed_size;
	struct page **pages;
	unsigned long nr_pages;
284
	int compress_type;
285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301
	struct list_head list;
};

struct async_cow {
	struct inode *inode;
	struct btrfs_root *root;
	struct page *locked_page;
	u64 start;
	u64 end;
	struct list_head extents;
	struct btrfs_work work;
};

static noinline int add_async_extent(struct async_cow *cow,
				     u64 start, u64 ram_size,
				     u64 compressed_size,
				     struct page **pages,
302 303
				     unsigned long nr_pages,
				     int compress_type)
304 305 306 307
{
	struct async_extent *async_extent;

	async_extent = kmalloc(sizeof(*async_extent), GFP_NOFS);
308
	BUG_ON(!async_extent); /* -ENOMEM */
309 310 311 312 313
	async_extent->start = start;
	async_extent->ram_size = ram_size;
	async_extent->compressed_size = compressed_size;
	async_extent->pages = pages;
	async_extent->nr_pages = nr_pages;
314
	async_extent->compress_type = compress_type;
315 316 317 318
	list_add_tail(&async_extent->list, &cow->extents);
	return 0;
}

C
Chris Mason 已提交
319
/*
320 321 322
 * we create compressed extents in two phases.  The first
 * phase compresses a range of pages that have already been
 * locked (both pages and state bits are locked).
C
Chris Mason 已提交
323
 *
324 325 326 327 328
 * This is done inside an ordered work queue, and the compression
 * is spread across many cpus.  The actual IO submission is step
 * two, and the ordered work queue takes care of making sure that
 * happens in the same order things were put onto the queue by
 * writepages and friends.
C
Chris Mason 已提交
329
 *
330 331 332
 * If this code finds it can't get good compression, it puts an
 * entry onto the work queue to write the uncompressed bytes.  This
 * makes sure that both compressed inodes and uncompressed inodes
333 334
 * are written in the same order that the flusher thread sent them
 * down.
C
Chris Mason 已提交
335
 */
336 337 338 339 340
static noinline int compress_file_range(struct inode *inode,
					struct page *locked_page,
					u64 start, u64 end,
					struct async_cow *async_cow,
					int *num_added)
341 342 343
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
344 345
	u64 num_bytes;
	u64 blocksize = root->sectorsize;
C
Chris Mason 已提交
346
	u64 actual_end;
347
	u64 isize = i_size_read(inode);
348
	int ret = 0;
C
Chris Mason 已提交
349 350 351 352 353 354
	struct page **pages = NULL;
	unsigned long nr_pages;
	unsigned long nr_pages_ret = 0;
	unsigned long total_compressed = 0;
	unsigned long total_in = 0;
	unsigned long max_compressed = 128 * 1024;
355
	unsigned long max_uncompressed = 128 * 1024;
C
Chris Mason 已提交
356 357
	int i;
	int will_compress;
358
	int compress_type = root->fs_info->compress_type;
359
	int redirty = 0;
360

361 362 363
	/* if this is a small write inside eof, kick off a defrag */
	if ((end - start + 1) < 16 * 1024 &&
	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
C
Chris Mason 已提交
364 365
		btrfs_add_inode_defrag(NULL, inode);

366
	actual_end = min_t(u64, isize, end + 1);
C
Chris Mason 已提交
367 368 369 370
again:
	will_compress = 0;
	nr_pages = (end >> PAGE_CACHE_SHIFT) - (start >> PAGE_CACHE_SHIFT) + 1;
	nr_pages = min(nr_pages, (128 * 1024UL) / PAGE_CACHE_SIZE);
371

372 373 374 375 376 377 378 379 380 381 382 383 384
	/*
	 * we don't want to send crud past the end of i_size through
	 * compression, that's just a waste of CPU time.  So, if the
	 * end of the file is before the start of our current
	 * requested range of bytes, we bail out to the uncompressed
	 * cleanup code that can deal with all of this.
	 *
	 * It isn't really the fastest way to fix things, but this is a
	 * very uncommon corner.
	 */
	if (actual_end <= start)
		goto cleanup_and_bail_uncompressed;

C
Chris Mason 已提交
385 386 387 388
	total_compressed = actual_end - start;

	/* we want to make sure that amount of ram required to uncompress
	 * an extent is reasonable, so we limit the total size in ram
389 390 391 392 393 394 395
	 * of a compressed extent to 128k.  This is a crucial number
	 * because it also controls how easily we can spread reads across
	 * cpus for decompression.
	 *
	 * We also want to make sure the amount of IO required to do
	 * a random read is reasonably small, so we limit the size of
	 * a compressed extent to 128k.
C
Chris Mason 已提交
396 397
	 */
	total_compressed = min(total_compressed, max_uncompressed);
398
	num_bytes = ALIGN(end - start + 1, blocksize);
399
	num_bytes = max(blocksize,  num_bytes);
C
Chris Mason 已提交
400 401
	total_in = 0;
	ret = 0;
402

403 404 405 406
	/*
	 * we do compression for mount -o compress and when the
	 * inode has not been flagged as nocompress.  This flag can
	 * change at any time if we discover bad compression ratios.
C
Chris Mason 已提交
407
	 */
408
	if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS) &&
C
Chris Mason 已提交
409
	    (btrfs_test_opt(root, COMPRESS) ||
410 411
	     (BTRFS_I(inode)->force_compress) ||
	     (BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS))) {
C
Chris Mason 已提交
412
		WARN_ON(pages);
413
		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
414 415 416 417
		if (!pages) {
			/* just bail out to the uncompressed code */
			goto cont;
		}
C
Chris Mason 已提交
418

419 420 421
		if (BTRFS_I(inode)->force_compress)
			compress_type = BTRFS_I(inode)->force_compress;

422 423 424 425 426 427 428 429 430 431 432
		/*
		 * we need to call clear_page_dirty_for_io on each
		 * page in the range.  Otherwise applications with the file
		 * mmap'd can wander in and change the page contents while
		 * we are compressing them.
		 *
		 * If the compression fails for any reason, we set the pages
		 * dirty again later on.
		 */
		extent_range_clear_dirty_for_io(inode, start, end);
		redirty = 1;
433 434 435 436 437 438 439
		ret = btrfs_compress_pages(compress_type,
					   inode->i_mapping, start,
					   total_compressed, pages,
					   nr_pages, &nr_pages_ret,
					   &total_in,
					   &total_compressed,
					   max_compressed);
C
Chris Mason 已提交
440 441 442 443 444 445 446 447 448 449 450

		if (!ret) {
			unsigned long offset = total_compressed &
				(PAGE_CACHE_SIZE - 1);
			struct page *page = pages[nr_pages_ret - 1];
			char *kaddr;

			/* zero the tail end of the last page, we might be
			 * sending it down to disk
			 */
			if (offset) {
451
				kaddr = kmap_atomic(page);
C
Chris Mason 已提交
452 453
				memset(kaddr + offset, 0,
				       PAGE_CACHE_SIZE - offset);
454
				kunmap_atomic(kaddr);
C
Chris Mason 已提交
455 456 457 458
			}
			will_compress = 1;
		}
	}
459
cont:
C
Chris Mason 已提交
460
	if (start == 0) {
461
		trans = btrfs_join_transaction(root);
462 463 464 465 466
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
			trans = NULL;
			goto cleanup_and_out;
		}
467
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
468

C
Chris Mason 已提交
469
		/* lets try to make an inline extent */
470
		if (ret || total_in < (actual_end - start)) {
C
Chris Mason 已提交
471
			/* we didn't compress the entire range, try
472
			 * to make an uncompressed inline extent.
C
Chris Mason 已提交
473 474
			 */
			ret = cow_file_range_inline(trans, root, inode,
475
						    start, end, 0, 0, NULL);
C
Chris Mason 已提交
476
		} else {
477
			/* try making a compressed inline extent */
C
Chris Mason 已提交
478 479
			ret = cow_file_range_inline(trans, root, inode,
						    start, end,
480 481
						    total_compressed,
						    compress_type, pages);
C
Chris Mason 已提交
482
		}
483
		if (ret <= 0) {
484
			/*
485 486 487
			 * inline extent creation worked or returned error,
			 * we don't need to create any more async work items.
			 * Unlock and free up our temp pages.
488
			 */
C
Chris Mason 已提交
489
			extent_clear_unlock_delalloc(inode,
490 491 492
			     &BTRFS_I(inode)->io_tree,
			     start, end, NULL,
			     EXTENT_CLEAR_UNLOCK_PAGE | EXTENT_CLEAR_DIRTY |
493
			     EXTENT_CLEAR_DELALLOC |
494
			     EXTENT_SET_WRITEBACK | EXTENT_END_WRITEBACK);
495 496

			btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
497 498
			goto free_pages_out;
		}
499
		btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
500 501 502 503 504 505 506 507
	}

	if (will_compress) {
		/*
		 * we aren't doing an inline extent round the compressed size
		 * up to a block size boundary so the allocator does sane
		 * things
		 */
508
		total_compressed = ALIGN(total_compressed, blocksize);
C
Chris Mason 已提交
509 510 511 512 513

		/*
		 * one last check to make sure the compression is really a
		 * win, compare the page count read with the blocks on disk
		 */
514
		total_in = ALIGN(total_in, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
515 516 517 518 519 520 521 522 523 524 525 526
		if (total_compressed >= total_in) {
			will_compress = 0;
		} else {
			num_bytes = total_in;
		}
	}
	if (!will_compress && pages) {
		/*
		 * the compression code ran but failed to make things smaller,
		 * free any pages it allocated and our page pointer array
		 */
		for (i = 0; i < nr_pages_ret; i++) {
C
Chris Mason 已提交
527
			WARN_ON(pages[i]->mapping);
C
Chris Mason 已提交
528 529 530 531 532 533 534 535
			page_cache_release(pages[i]);
		}
		kfree(pages);
		pages = NULL;
		total_compressed = 0;
		nr_pages_ret = 0;

		/* flag the file so we don't compress in the future */
C
Chris Mason 已提交
536 537
		if (!btrfs_test_opt(root, FORCE_COMPRESS) &&
		    !(BTRFS_I(inode)->force_compress)) {
C
Chris Mason 已提交
538
			BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
C
Chris Mason 已提交
539
		}
C
Chris Mason 已提交
540
	}
541 542
	if (will_compress) {
		*num_added += 1;
C
Chris Mason 已提交
543

544 545 546 547 548
		/* the async work queues will take care of doing actual
		 * allocation on disk for these compressed pages,
		 * and will submit them to the elevator.
		 */
		add_async_extent(async_cow, start, num_bytes,
549 550
				 total_compressed, pages, nr_pages_ret,
				 compress_type);
551

552
		if (start + num_bytes < end) {
553 554 555 556 557 558
			start += num_bytes;
			pages = NULL;
			cond_resched();
			goto again;
		}
	} else {
559
cleanup_and_bail_uncompressed:
560 561 562 563 564 565 566 567 568 569 570 571
		/*
		 * No compression, but we still need to write the pages in
		 * the file we've been given so far.  redirty the locked
		 * page if it corresponds to our extent and set things up
		 * for the async work queue to run cow_file_range to do
		 * the normal delalloc dance
		 */
		if (page_offset(locked_page) >= start &&
		    page_offset(locked_page) <= end) {
			__set_page_dirty_nobuffers(locked_page);
			/* unlocked later on in the async handlers */
		}
572 573
		if (redirty)
			extent_range_redirty_for_io(inode, start, end);
574 575
		add_async_extent(async_cow, start, end - start + 1,
				 0, NULL, 0, BTRFS_COMPRESS_NONE);
576 577
		*num_added += 1;
	}
578

579
out:
580
	return ret;
581 582 583 584 585 586

free_pages_out:
	for (i = 0; i < nr_pages_ret; i++) {
		WARN_ON(pages[i]->mapping);
		page_cache_release(pages[i]);
	}
C
Chris Mason 已提交
587
	kfree(pages);
588 589

	goto out;
590 591 592 593 594 595 596 597 598 599 600 601 602 603

cleanup_and_out:
	extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
	if (!trans || IS_ERR(trans))
		btrfs_error(root->fs_info, ret, "Failed to join transaction");
	else
		btrfs_abort_transaction(trans, root, ret);
	goto free_pages_out;
604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
}

/*
 * phase two of compressed writeback.  This is the ordered portion
 * of the code, which only gets called in the order the work was
 * queued.  We walk all the async extents created by compress_file_range
 * and send them down to the disk.
 */
static noinline int submit_compressed_extents(struct inode *inode,
					      struct async_cow *async_cow)
{
	struct async_extent *async_extent;
	u64 alloc_hint = 0;
	struct btrfs_trans_handle *trans;
	struct btrfs_key ins;
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_io_tree *io_tree;
623
	int ret = 0;
624 625 626 627

	if (list_empty(&async_cow->extents))
		return 0;

628
again:
C
Chris Mason 已提交
629
	while (!list_empty(&async_cow->extents)) {
630 631 632
		async_extent = list_entry(async_cow->extents.next,
					  struct async_extent, list);
		list_del(&async_extent->list);
C
Chris Mason 已提交
633

634 635
		io_tree = &BTRFS_I(inode)->io_tree;

636
retry:
637 638 639 640 641 642
		/* did the compression code fall back to uncompressed IO? */
		if (!async_extent->pages) {
			int page_started = 0;
			unsigned long nr_written = 0;

			lock_extent(io_tree, async_extent->start,
643
					 async_extent->start +
644
					 async_extent->ram_size - 1);
645 646

			/* allocate blocks */
647 648 649 650 651
			ret = cow_file_range(inode, async_cow->locked_page,
					     async_extent->start,
					     async_extent->start +
					     async_extent->ram_size - 1,
					     &page_started, &nr_written, 0);
652

653 654
			/* JDM XXX */

655 656 657 658 659 660
			/*
			 * if page_started, cow_file_range inserted an
			 * inline extent and took care of all the unlocking
			 * and IO for us.  Otherwise, we need to submit
			 * all those pages down to the drive.
			 */
661
			if (!page_started && !ret)
662 663
				extent_write_locked_range(io_tree,
						  inode, async_extent->start,
C
Chris Mason 已提交
664
						  async_extent->start +
665 666 667
						  async_extent->ram_size - 1,
						  btrfs_get_extent,
						  WB_SYNC_ALL);
668 669
			else if (ret)
				unlock_page(async_cow->locked_page);
670 671 672 673 674 675
			kfree(async_extent);
			cond_resched();
			continue;
		}

		lock_extent(io_tree, async_extent->start,
676
			    async_extent->start + async_extent->ram_size - 1);
677

678
		trans = btrfs_join_transaction(root);
679 680 681 682 683
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
		} else {
			trans->block_rsv = &root->fs_info->delalloc_block_rsv;
			ret = btrfs_reserve_extent(trans, root,
684 685
					   async_extent->compressed_size,
					   async_extent->compressed_size,
686
					   0, alloc_hint, &ins, 1);
687
			if (ret && ret != -ENOSPC)
688 689 690
				btrfs_abort_transaction(trans, root, ret);
			btrfs_end_transaction(trans, root);
		}
691

692 693
		if (ret) {
			int i;
694

695 696 697 698 699 700 701
			for (i = 0; i < async_extent->nr_pages; i++) {
				WARN_ON(async_extent->pages[i]->mapping);
				page_cache_release(async_extent->pages[i]);
			}
			kfree(async_extent->pages);
			async_extent->nr_pages = 0;
			async_extent->pages = NULL;
702

703 704
			if (ret == -ENOSPC)
				goto retry;
705
			goto out_free;
706 707
		}

708 709 710 711 712 713 714 715
		/*
		 * here we're doing allocation and writeback of the
		 * compressed pages
		 */
		btrfs_drop_extent_cache(inode, async_extent->start,
					async_extent->start +
					async_extent->ram_size - 1, 0);

716
		em = alloc_extent_map();
717 718
		if (!em) {
			ret = -ENOMEM;
719
			goto out_free_reserve;
720
		}
721 722
		em->start = async_extent->start;
		em->len = async_extent->ram_size;
723
		em->orig_start = em->start;
724 725
		em->mod_start = em->start;
		em->mod_len = em->len;
C
Chris Mason 已提交
726

727 728
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
729
		em->orig_block_len = ins.offset;
J
Josef Bacik 已提交
730
		em->ram_bytes = async_extent->ram_size;
731
		em->bdev = root->fs_info->fs_devices->latest_bdev;
732
		em->compress_type = async_extent->compress_type;
733 734
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
		set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
735
		em->generation = -1;
736

C
Chris Mason 已提交
737
		while (1) {
738
			write_lock(&em_tree->lock);
J
Josef Bacik 已提交
739
			ret = add_extent_mapping(em_tree, em, 1);
740
			write_unlock(&em_tree->lock);
741 742 743 744 745 746 747 748 749
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, async_extent->start,
						async_extent->start +
						async_extent->ram_size - 1, 0);
		}

750 751 752
		if (ret)
			goto out_free_reserve;

753 754 755 756 757 758 759
		ret = btrfs_add_ordered_extent_compress(inode,
						async_extent->start,
						ins.objectid,
						async_extent->ram_size,
						ins.offset,
						BTRFS_ORDERED_COMPRESSED,
						async_extent->compress_type);
760 761
		if (ret)
			goto out_free_reserve;
762 763 764 765 766

		/*
		 * clear dirty, set writeback and unlock the pages.
		 */
		extent_clear_unlock_delalloc(inode,
767 768 769 770 771 772
				&BTRFS_I(inode)->io_tree,
				async_extent->start,
				async_extent->start +
				async_extent->ram_size - 1,
				NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK |
773
				EXTENT_CLEAR_DELALLOC |
774
				EXTENT_CLEAR_DIRTY | EXTENT_SET_WRITEBACK);
775 776

		ret = btrfs_submit_compressed_write(inode,
C
Chris Mason 已提交
777 778 779 780 781
				    async_extent->start,
				    async_extent->ram_size,
				    ins.objectid,
				    ins.offset, async_extent->pages,
				    async_extent->nr_pages);
782 783
		alloc_hint = ins.objectid + ins.offset;
		kfree(async_extent);
784 785
		if (ret)
			goto out;
786 787
		cond_resched();
	}
788 789 790
	ret = 0;
out:
	return ret;
791 792
out_free_reserve:
	btrfs_free_reserved_extent(root, ins.objectid, ins.offset);
793
out_free:
794 795 796 797 798 799 800 801 802 803
	extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
				     async_extent->start,
				     async_extent->start +
				     async_extent->ram_size - 1,
				     NULL, EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_UNLOCK |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
804
	kfree(async_extent);
805
	goto again;
806 807
}

808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839
static u64 get_extent_allocation_hint(struct inode *inode, u64 start,
				      u64 num_bytes)
{
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
	u64 alloc_hint = 0;

	read_lock(&em_tree->lock);
	em = search_extent_mapping(em_tree, start, num_bytes);
	if (em) {
		/*
		 * if block start isn't an actual block number then find the
		 * first block in this inode and use that as a hint.  If that
		 * block is also bogus then just don't worry about it.
		 */
		if (em->block_start >= EXTENT_MAP_LAST_BYTE) {
			free_extent_map(em);
			em = search_extent_mapping(em_tree, 0, 0);
			if (em && em->block_start < EXTENT_MAP_LAST_BYTE)
				alloc_hint = em->block_start;
			if (em)
				free_extent_map(em);
		} else {
			alloc_hint = em->block_start;
			free_extent_map(em);
		}
	}
	read_unlock(&em_tree->lock);

	return alloc_hint;
}

840 841 842 843 844 845 846 847 848 849 850 851 852
/*
 * when extent_io.c finds a delayed allocation range in the file,
 * the call backs end up in this code.  The basic idea is to
 * allocate extents on disk for the range, and create ordered data structs
 * in ram to track those extents.
 *
 * locked_page is the page that writepage had locked already.  We use
 * it to make sure we don't do extra locks or unlocks.
 *
 * *page_started is set to one if we unlock locked_page and do everything
 * required to start IO on it.  It may be clean and already done with
 * IO when we return.
 */
853 854 855 856 857 858 859
static noinline int __cow_file_range(struct btrfs_trans_handle *trans,
				     struct inode *inode,
				     struct btrfs_root *root,
				     struct page *locked_page,
				     u64 start, u64 end, int *page_started,
				     unsigned long *nr_written,
				     int unlock)
860 861 862 863 864 865 866 867 868 869 870 871
{
	u64 alloc_hint = 0;
	u64 num_bytes;
	unsigned long ram_size;
	u64 disk_num_bytes;
	u64 cur_alloc_size;
	u64 blocksize = root->sectorsize;
	struct btrfs_key ins;
	struct extent_map *em;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	int ret = 0;

872
	BUG_ON(btrfs_is_free_space_inode(inode));
873

874
	num_bytes = ALIGN(end - start + 1, blocksize);
875 876 877
	num_bytes = max(blocksize,  num_bytes);
	disk_num_bytes = num_bytes;

C
Chris Mason 已提交
878
	/* if this is a small write inside eof, kick off defrag */
879 880
	if (num_bytes < 64 * 1024 &&
	    (start > 0 || end + 1 < BTRFS_I(inode)->disk_i_size))
C
Chris Mason 已提交
881 882
		btrfs_add_inode_defrag(trans, inode);

883 884 885
	if (start == 0) {
		/* lets try to make an inline extent */
		ret = cow_file_range_inline(trans, root, inode,
886
					    start, end, 0, 0, NULL);
887 888
		if (ret == 0) {
			extent_clear_unlock_delalloc(inode,
889 890 891 892 893 894 895 896
				     &BTRFS_I(inode)->io_tree,
				     start, end, NULL,
				     EXTENT_CLEAR_UNLOCK_PAGE |
				     EXTENT_CLEAR_UNLOCK |
				     EXTENT_CLEAR_DELALLOC |
				     EXTENT_CLEAR_DIRTY |
				     EXTENT_SET_WRITEBACK |
				     EXTENT_END_WRITEBACK);
897

898 899 900 901
			*nr_written = *nr_written +
			     (end - start + PAGE_CACHE_SIZE) / PAGE_CACHE_SIZE;
			*page_started = 1;
			goto out;
902 903 904
		} else if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_unlock;
905 906 907 908
		}
	}

	BUG_ON(disk_num_bytes >
909
	       btrfs_super_total_bytes(root->fs_info->super_copy));
910

911
	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
912 913
	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);

C
Chris Mason 已提交
914
	while (disk_num_bytes > 0) {
915 916
		unsigned long op;

917
		cur_alloc_size = disk_num_bytes;
918
		ret = btrfs_reserve_extent(trans, root, cur_alloc_size,
919
					   root->sectorsize, 0, alloc_hint,
920
					   &ins, 1);
921 922 923 924
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_unlock;
		}
C
Chris Mason 已提交
925

926
		em = alloc_extent_map();
927 928
		if (!em) {
			ret = -ENOMEM;
929
			goto out_reserve;
930
		}
931
		em->start = start;
932
		em->orig_start = em->start;
933 934
		ram_size = ins.offset;
		em->len = ins.offset;
935 936
		em->mod_start = em->start;
		em->mod_len = em->len;
C
Chris Mason 已提交
937

938
		em->block_start = ins.objectid;
C
Chris Mason 已提交
939
		em->block_len = ins.offset;
940
		em->orig_block_len = ins.offset;
J
Josef Bacik 已提交
941
		em->ram_bytes = ram_size;
942
		em->bdev = root->fs_info->fs_devices->latest_bdev;
943
		set_bit(EXTENT_FLAG_PINNED, &em->flags);
944
		em->generation = -1;
C
Chris Mason 已提交
945

C
Chris Mason 已提交
946
		while (1) {
947
			write_lock(&em_tree->lock);
J
Josef Bacik 已提交
948
			ret = add_extent_mapping(em_tree, em, 1);
949
			write_unlock(&em_tree->lock);
950 951 952 953 954
			if (ret != -EEXIST) {
				free_extent_map(em);
				break;
			}
			btrfs_drop_extent_cache(inode, start,
C
Chris Mason 已提交
955
						start + ram_size - 1, 0);
956
		}
957 958
		if (ret)
			goto out_reserve;
959

960
		cur_alloc_size = ins.offset;
961
		ret = btrfs_add_ordered_extent(inode, start, ins.objectid,
962
					       ram_size, cur_alloc_size, 0);
963 964
		if (ret)
			goto out_reserve;
C
Chris Mason 已提交
965

966 967 968 969
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, start,
						      cur_alloc_size);
970 971
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
972
				goto out_reserve;
973
			}
974 975
		}

C
Chris Mason 已提交
976
		if (disk_num_bytes < cur_alloc_size)
977
			break;
C
Chris Mason 已提交
978

C
Chris Mason 已提交
979 980 981
		/* we're not doing compressed IO, don't unlock the first
		 * page (which the caller expects to stay locked), don't
		 * clear any dirty bits and don't set any writeback bits
982 983 984
		 *
		 * Do set the Private2 bit so we know this page was properly
		 * setup for writepage
C
Chris Mason 已提交
985
		 */
986 987 988 989
		op = unlock ? EXTENT_CLEAR_UNLOCK_PAGE : 0;
		op |= EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
			EXTENT_SET_PRIVATE2;

C
Chris Mason 已提交
990 991
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
					     start, start + ram_size - 1,
992
					     locked_page, op);
C
Chris Mason 已提交
993
		disk_num_bytes -= cur_alloc_size;
994 995 996
		num_bytes -= cur_alloc_size;
		alloc_hint = ins.objectid + ins.offset;
		start += cur_alloc_size;
997
	}
998
out:
999
	return ret;
1000

1001 1002
out_reserve:
	btrfs_free_reserved_extent(root, ins.objectid, ins.offset);
1003 1004 1005
out_unlock:
	extent_clear_unlock_delalloc(inode,
		     &BTRFS_I(inode)->io_tree,
1006
		     start, end, locked_page,
1007 1008 1009 1010 1011 1012 1013 1014
		     EXTENT_CLEAR_UNLOCK_PAGE |
		     EXTENT_CLEAR_UNLOCK |
		     EXTENT_CLEAR_DELALLOC |
		     EXTENT_CLEAR_DIRTY |
		     EXTENT_SET_WRITEBACK |
		     EXTENT_END_WRITEBACK);

	goto out;
1015
}
C
Chris Mason 已提交
1016

1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049
static noinline int cow_file_range(struct inode *inode,
				   struct page *locked_page,
				   u64 start, u64 end, int *page_started,
				   unsigned long *nr_written,
				   int unlock)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	trans = btrfs_join_transaction(root);
	if (IS_ERR(trans)) {
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
		return PTR_ERR(trans);
	}
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;

	ret = __cow_file_range(trans, inode, root, locked_page, start, end,
			       page_started, nr_written, unlock);

	btrfs_end_transaction(trans, root);

	return ret;
}

1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061
/*
 * work queue call back to started compression on a file and pages
 */
static noinline void async_cow_start(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	int num_added = 0;
	async_cow = container_of(work, struct async_cow, work);

	compress_file_range(async_cow->inode, async_cow->locked_page,
			    async_cow->start, async_cow->end, async_cow,
			    &num_added);
1062
	if (num_added == 0) {
1063
		btrfs_add_delayed_iput(async_cow->inode);
1064
		async_cow->inode = NULL;
1065
	}
1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082
}

/*
 * work queue call back to submit previously compressed pages
 */
static noinline void async_cow_submit(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	struct btrfs_root *root;
	unsigned long nr_pages;

	async_cow = container_of(work, struct async_cow, work);

	root = async_cow->root;
	nr_pages = (async_cow->end - async_cow->start + PAGE_CACHE_SIZE) >>
		PAGE_CACHE_SHIFT;

1083
	if (atomic_sub_return(nr_pages, &root->fs_info->async_delalloc_pages) <
1084
	    5 * 1024 * 1024 &&
1085 1086 1087
	    waitqueue_active(&root->fs_info->async_submit_wait))
		wake_up(&root->fs_info->async_submit_wait);

C
Chris Mason 已提交
1088
	if (async_cow->inode)
1089 1090
		submit_compressed_extents(async_cow->inode, async_cow);
}
C
Chris Mason 已提交
1091

1092 1093 1094 1095
static noinline void async_cow_free(struct btrfs_work *work)
{
	struct async_cow *async_cow;
	async_cow = container_of(work, struct async_cow, work);
1096
	if (async_cow->inode)
1097
		btrfs_add_delayed_iput(async_cow->inode);
1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108
	kfree(async_cow);
}

static int cow_file_range_async(struct inode *inode, struct page *locked_page,
				u64 start, u64 end, int *page_started,
				unsigned long *nr_written)
{
	struct async_cow *async_cow;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	unsigned long nr_pages;
	u64 cur_end;
1109
	int limit = 10 * 1024 * 1024;
1110

1111 1112
	clear_extent_bit(&BTRFS_I(inode)->io_tree, start, end, EXTENT_LOCKED,
			 1, 0, NULL, GFP_NOFS);
C
Chris Mason 已提交
1113
	while (start < end) {
1114
		async_cow = kmalloc(sizeof(*async_cow), GFP_NOFS);
1115
		BUG_ON(!async_cow); /* -ENOMEM */
1116
		async_cow->inode = igrab(inode);
1117 1118 1119 1120
		async_cow->root = root;
		async_cow->locked_page = locked_page;
		async_cow->start = start;

1121
		if (BTRFS_I(inode)->flags & BTRFS_INODE_NOCOMPRESS)
1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146
			cur_end = end;
		else
			cur_end = min(end, start + 512 * 1024 - 1);

		async_cow->end = cur_end;
		INIT_LIST_HEAD(&async_cow->extents);

		async_cow->work.func = async_cow_start;
		async_cow->work.ordered_func = async_cow_submit;
		async_cow->work.ordered_free = async_cow_free;
		async_cow->work.flags = 0;

		nr_pages = (cur_end - start + PAGE_CACHE_SIZE) >>
			PAGE_CACHE_SHIFT;
		atomic_add(nr_pages, &root->fs_info->async_delalloc_pages);

		btrfs_queue_worker(&root->fs_info->delalloc_workers,
				   &async_cow->work);

		if (atomic_read(&root->fs_info->async_delalloc_pages) > limit) {
			wait_event(root->fs_info->async_submit_wait,
			   (atomic_read(&root->fs_info->async_delalloc_pages) <
			    limit));
		}

C
Chris Mason 已提交
1147
		while (atomic_read(&root->fs_info->async_submit_draining) &&
1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158
		      atomic_read(&root->fs_info->async_delalloc_pages)) {
			wait_event(root->fs_info->async_submit_wait,
			  (atomic_read(&root->fs_info->async_delalloc_pages) ==
			   0));
		}

		*nr_written += nr_pages;
		start = cur_end + 1;
	}
	*page_started = 1;
	return 0;
1159 1160
}

C
Chris Mason 已提交
1161
static noinline int csum_exist_in_range(struct btrfs_root *root,
1162 1163 1164 1165 1166 1167
					u64 bytenr, u64 num_bytes)
{
	int ret;
	struct btrfs_ordered_sum *sums;
	LIST_HEAD(list);

Y
Yan Zheng 已提交
1168
	ret = btrfs_lookup_csums_range(root->fs_info->csum_root, bytenr,
A
Arne Jansen 已提交
1169
				       bytenr + num_bytes - 1, &list, 0);
1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180
	if (ret == 0 && list_empty(&list))
		return 0;

	while (!list_empty(&list)) {
		sums = list_entry(list.next, struct btrfs_ordered_sum, list);
		list_del(&sums->list);
		kfree(sums);
	}
	return 1;
}

C
Chris Mason 已提交
1181 1182 1183 1184 1185 1186 1187
/*
 * when nowcow writeback call back.  This checks for snapshots or COW copies
 * of the extents that exist in the file, and COWs the file as required.
 *
 * If no cow copies or snapshots exist, we write directly to the existing
 * blocks on disk
 */
1188 1189
static noinline int run_delalloc_nocow(struct inode *inode,
				       struct page *locked_page,
1190 1191
			      u64 start, u64 end, int *page_started, int force,
			      unsigned long *nr_written)
1192 1193
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1194
	struct btrfs_trans_handle *trans;
1195 1196
	struct extent_buffer *leaf;
	struct btrfs_path *path;
Y
Yan Zheng 已提交
1197
	struct btrfs_file_extent_item *fi;
1198
	struct btrfs_key found_key;
Y
Yan Zheng 已提交
1199 1200 1201
	u64 cow_start;
	u64 cur_offset;
	u64 extent_end;
1202
	u64 extent_offset;
Y
Yan Zheng 已提交
1203 1204
	u64 disk_bytenr;
	u64 num_bytes;
1205
	u64 disk_num_bytes;
J
Josef Bacik 已提交
1206
	u64 ram_bytes;
Y
Yan Zheng 已提交
1207
	int extent_type;
1208
	int ret, err;
Y
Yan Zheng 已提交
1209
	int type;
Y
Yan Zheng 已提交
1210 1211
	int nocow;
	int check_prev = 1;
1212
	bool nolock;
L
Li Zefan 已提交
1213
	u64 ino = btrfs_ino(inode);
1214 1215

	path = btrfs_alloc_path();
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
	if (!path) {
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
1226
		return -ENOMEM;
1227
	}
1228

1229
	nolock = btrfs_is_free_space_inode(inode);
1230 1231

	if (nolock)
1232
		trans = btrfs_join_transaction_nolock(root);
1233
	else
1234
		trans = btrfs_join_transaction(root);
C
Chris Mason 已提交
1235

1236
	if (IS_ERR(trans)) {
1237 1238 1239 1240 1241 1242 1243 1244 1245
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     start, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);
1246 1247 1248 1249
		btrfs_free_path(path);
		return PTR_ERR(trans);
	}

1250
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
1251

Y
Yan Zheng 已提交
1252 1253 1254
	cow_start = (u64)-1;
	cur_offset = start;
	while (1) {
L
Li Zefan 已提交
1255
		ret = btrfs_lookup_file_extent(trans, root, path, ino,
Y
Yan Zheng 已提交
1256
					       cur_offset, 0);
1257 1258 1259 1260
		if (ret < 0) {
			btrfs_abort_transaction(trans, root, ret);
			goto error;
		}
Y
Yan Zheng 已提交
1261 1262 1263 1264
		if (ret > 0 && path->slots[0] > 0 && check_prev) {
			leaf = path->nodes[0];
			btrfs_item_key_to_cpu(leaf, &found_key,
					      path->slots[0] - 1);
L
Li Zefan 已提交
1265
			if (found_key.objectid == ino &&
Y
Yan Zheng 已提交
1266 1267 1268 1269 1270 1271 1272 1273
			    found_key.type == BTRFS_EXTENT_DATA_KEY)
				path->slots[0]--;
		}
		check_prev = 0;
next_slot:
		leaf = path->nodes[0];
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
1274 1275 1276 1277
			if (ret < 0) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
Y
Yan Zheng 已提交
1278 1279 1280 1281
			if (ret > 0)
				break;
			leaf = path->nodes[0];
		}
1282

Y
Yan Zheng 已提交
1283 1284
		nocow = 0;
		disk_bytenr = 0;
1285
		num_bytes = 0;
Y
Yan Zheng 已提交
1286 1287
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

L
Li Zefan 已提交
1288
		if (found_key.objectid > ino ||
Y
Yan Zheng 已提交
1289 1290 1291 1292 1293 1294
		    found_key.type > BTRFS_EXTENT_DATA_KEY ||
		    found_key.offset > end)
			break;

		if (found_key.offset > cur_offset) {
			extent_end = found_key.offset;
1295
			extent_type = 0;
Y
Yan Zheng 已提交
1296 1297 1298 1299 1300 1301 1302
			goto out_check;
		}

		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		extent_type = btrfs_file_extent_type(leaf, fi);

J
Josef Bacik 已提交
1303
		ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);
Y
Yan Zheng 已提交
1304 1305
		if (extent_type == BTRFS_FILE_EXTENT_REG ||
		    extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
1306
			disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
1307
			extent_offset = btrfs_file_extent_offset(leaf, fi);
Y
Yan Zheng 已提交
1308 1309
			extent_end = found_key.offset +
				btrfs_file_extent_num_bytes(leaf, fi);
1310 1311
			disk_num_bytes =
				btrfs_file_extent_disk_num_bytes(leaf, fi);
Y
Yan Zheng 已提交
1312 1313 1314 1315
			if (extent_end <= start) {
				path->slots[0]++;
				goto next_slot;
			}
1316 1317
			if (disk_bytenr == 0)
				goto out_check;
Y
Yan Zheng 已提交
1318 1319 1320 1321
			if (btrfs_file_extent_compression(leaf, fi) ||
			    btrfs_file_extent_encryption(leaf, fi) ||
			    btrfs_file_extent_other_encoding(leaf, fi))
				goto out_check;
Y
Yan Zheng 已提交
1322 1323
			if (extent_type == BTRFS_FILE_EXTENT_REG && !force)
				goto out_check;
1324
			if (btrfs_extent_readonly(root, disk_bytenr))
Y
Yan Zheng 已提交
1325
				goto out_check;
L
Li Zefan 已提交
1326
			if (btrfs_cross_ref_exist(trans, root, ino,
1327 1328
						  found_key.offset -
						  extent_offset, disk_bytenr))
1329
				goto out_check;
1330
			disk_bytenr += extent_offset;
1331 1332 1333 1334 1335 1336 1337 1338 1339
			disk_bytenr += cur_offset - found_key.offset;
			num_bytes = min(end + 1, extent_end) - cur_offset;
			/*
			 * force cow if csum exists in the range.
			 * this ensure that csum for a given extent are
			 * either valid or do not exist.
			 */
			if (csum_exist_in_range(root, disk_bytenr, num_bytes))
				goto out_check;
Y
Yan Zheng 已提交
1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360
			nocow = 1;
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
			extent_end = found_key.offset +
				btrfs_file_extent_inline_len(leaf, fi);
			extent_end = ALIGN(extent_end, root->sectorsize);
		} else {
			BUG_ON(1);
		}
out_check:
		if (extent_end <= start) {
			path->slots[0]++;
			goto next_slot;
		}
		if (!nocow) {
			if (cow_start == (u64)-1)
				cow_start = cur_offset;
			cur_offset = extent_end;
			if (cur_offset > end)
				break;
			path->slots[0]++;
			goto next_slot;
1361 1362
		}

1363
		btrfs_release_path(path);
Y
Yan Zheng 已提交
1364
		if (cow_start != (u64)-1) {
1365 1366 1367
			ret = __cow_file_range(trans, inode, root, locked_page,
					       cow_start, found_key.offset - 1,
					       page_started, nr_written, 1);
1368 1369 1370 1371
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
Y
Yan Zheng 已提交
1372
			cow_start = (u64)-1;
1373
		}
Y
Yan Zheng 已提交
1374

Y
Yan Zheng 已提交
1375 1376 1377 1378
		if (extent_type == BTRFS_FILE_EXTENT_PREALLOC) {
			struct extent_map *em;
			struct extent_map_tree *em_tree;
			em_tree = &BTRFS_I(inode)->extent_tree;
1379
			em = alloc_extent_map();
1380
			BUG_ON(!em); /* -ENOMEM */
Y
Yan Zheng 已提交
1381
			em->start = cur_offset;
1382
			em->orig_start = found_key.offset - extent_offset;
Y
Yan Zheng 已提交
1383 1384 1385
			em->len = num_bytes;
			em->block_len = num_bytes;
			em->block_start = disk_bytenr;
1386
			em->orig_block_len = disk_num_bytes;
J
Josef Bacik 已提交
1387
			em->ram_bytes = ram_bytes;
Y
Yan Zheng 已提交
1388
			em->bdev = root->fs_info->fs_devices->latest_bdev;
1389 1390
			em->mod_start = em->start;
			em->mod_len = em->len;
Y
Yan Zheng 已提交
1391
			set_bit(EXTENT_FLAG_PINNED, &em->flags);
1392
			set_bit(EXTENT_FLAG_FILLING, &em->flags);
1393
			em->generation = -1;
Y
Yan Zheng 已提交
1394
			while (1) {
1395
				write_lock(&em_tree->lock);
J
Josef Bacik 已提交
1396
				ret = add_extent_mapping(em_tree, em, 1);
1397
				write_unlock(&em_tree->lock);
Y
Yan Zheng 已提交
1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
				if (ret != -EEXIST) {
					free_extent_map(em);
					break;
				}
				btrfs_drop_extent_cache(inode, em->start,
						em->start + em->len - 1, 0);
			}
			type = BTRFS_ORDERED_PREALLOC;
		} else {
			type = BTRFS_ORDERED_NOCOW;
		}
Y
Yan Zheng 已提交
1409 1410

		ret = btrfs_add_ordered_extent(inode, cur_offset, disk_bytenr,
Y
Yan Zheng 已提交
1411
					       num_bytes, num_bytes, type);
1412
		BUG_ON(ret); /* -ENOMEM */
1413

1414 1415 1416 1417
		if (root->root_key.objectid ==
		    BTRFS_DATA_RELOC_TREE_OBJECTID) {
			ret = btrfs_reloc_clone_csums(inode, cur_offset,
						      num_bytes);
1418 1419 1420 1421
			if (ret) {
				btrfs_abort_transaction(trans, root, ret);
				goto error;
			}
1422 1423
		}

Y
Yan Zheng 已提交
1424
		extent_clear_unlock_delalloc(inode, &BTRFS_I(inode)->io_tree,
1425 1426 1427 1428
				cur_offset, cur_offset + num_bytes - 1,
				locked_page, EXTENT_CLEAR_UNLOCK_PAGE |
				EXTENT_CLEAR_UNLOCK | EXTENT_CLEAR_DELALLOC |
				EXTENT_SET_PRIVATE2);
Y
Yan Zheng 已提交
1429 1430 1431
		cur_offset = extent_end;
		if (cur_offset > end)
			break;
1432
	}
1433
	btrfs_release_path(path);
Y
Yan Zheng 已提交
1434

1435
	if (cur_offset <= end && cow_start == (u64)-1) {
Y
Yan Zheng 已提交
1436
		cow_start = cur_offset;
1437 1438 1439
		cur_offset = end;
	}

Y
Yan Zheng 已提交
1440
	if (cow_start != (u64)-1) {
1441 1442 1443
		ret = __cow_file_range(trans, inode, root, locked_page,
				       cow_start, end,
				       page_started, nr_written, 1);
1444 1445 1446 1447
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			goto error;
		}
Y
Yan Zheng 已提交
1448 1449
	}

1450
error:
1451
	err = btrfs_end_transaction(trans, root);
1452 1453 1454
	if (!ret)
		ret = err;

1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
	if (ret && cur_offset < end)
		extent_clear_unlock_delalloc(inode,
			     &BTRFS_I(inode)->io_tree,
			     cur_offset, end, locked_page,
			     EXTENT_CLEAR_UNLOCK_PAGE |
			     EXTENT_CLEAR_UNLOCK |
			     EXTENT_CLEAR_DELALLOC |
			     EXTENT_CLEAR_DIRTY |
			     EXTENT_SET_WRITEBACK |
			     EXTENT_END_WRITEBACK);

1466
	btrfs_free_path(path);
1467
	return ret;
1468 1469
}

C
Chris Mason 已提交
1470 1471 1472
/*
 * extent_io.c call back to do delayed allocation processing
 */
C
Chris Mason 已提交
1473
static int run_delalloc_range(struct inode *inode, struct page *locked_page,
1474 1475
			      u64 start, u64 end, int *page_started,
			      unsigned long *nr_written)
1476 1477
{
	int ret;
1478
	struct btrfs_root *root = BTRFS_I(inode)->root;
1479

1480
	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) {
C
Chris Mason 已提交
1481
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1482
					 page_started, 1, nr_written);
1483
	} else if (BTRFS_I(inode)->flags & BTRFS_INODE_PREALLOC) {
Y
Yan Zheng 已提交
1484
		ret = run_delalloc_nocow(inode, locked_page, start, end,
C
Chris Mason 已提交
1485
					 page_started, 0, nr_written);
1486 1487 1488
	} else if (!btrfs_test_opt(root, COMPRESS) &&
		   !(BTRFS_I(inode)->force_compress) &&
		   !(BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS)) {
1489 1490
		ret = cow_file_range(inode, locked_page, start, end,
				      page_started, nr_written, 1);
1491 1492 1493
	} else {
		set_bit(BTRFS_INODE_HAS_ASYNC_EXTENT,
			&BTRFS_I(inode)->runtime_flags);
1494
		ret = cow_file_range_async(inode, locked_page, start, end,
C
Chris Mason 已提交
1495
					   page_started, nr_written);
1496
	}
1497 1498 1499
	return ret;
}

1500 1501
static void btrfs_split_extent_hook(struct inode *inode,
				    struct extent_state *orig, u64 split)
J
Josef Bacik 已提交
1502
{
1503
	/* not delalloc, ignore it */
J
Josef Bacik 已提交
1504
	if (!(orig->state & EXTENT_DELALLOC))
1505
		return;
J
Josef Bacik 已提交
1506

1507 1508 1509
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents++;
	spin_unlock(&BTRFS_I(inode)->lock);
J
Josef Bacik 已提交
1510 1511 1512 1513 1514 1515 1516 1517
}

/*
 * extent_io.c merge_extent_hook, used to track merged delayed allocation
 * extents so we can keep track of new extents that are just merged onto old
 * extents, such as when we are doing sequential writes, so we can properly
 * account for the metadata space we'll need.
 */
1518 1519 1520
static void btrfs_merge_extent_hook(struct inode *inode,
				    struct extent_state *new,
				    struct extent_state *other)
J
Josef Bacik 已提交
1521 1522 1523
{
	/* not delalloc, ignore it */
	if (!(other->state & EXTENT_DELALLOC))
1524
		return;
J
Josef Bacik 已提交
1525

1526 1527 1528
	spin_lock(&BTRFS_I(inode)->lock);
	BTRFS_I(inode)->outstanding_extents--;
	spin_unlock(&BTRFS_I(inode)->lock);
J
Josef Bacik 已提交
1529 1530
}

C
Chris Mason 已提交
1531 1532 1533 1534 1535
/*
 * extent_io.c set_bit_hook, used to track delayed allocation
 * bytes in this file, and to maintain the list of inodes that
 * have pending delalloc work to be done.
 */
1536
static void btrfs_set_bit_hook(struct inode *inode,
1537
			       struct extent_state *state, unsigned long *bits)
1538
{
J
Josef Bacik 已提交
1539

1540 1541
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
1542
	 * but in this case, we are only testing for the DELALLOC
1543 1544
	 * bit, which is only set or cleared with irqs on
	 */
1545
	if (!(state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1546
		struct btrfs_root *root = BTRFS_I(inode)->root;
1547
		u64 len = state->end + 1 - state->start;
1548
		bool do_list = !btrfs_is_free_space_inode(inode);
J
Josef Bacik 已提交
1549

1550
		if (*bits & EXTENT_FIRST_DELALLOC) {
1551
			*bits &= ~EXTENT_FIRST_DELALLOC;
1552 1553 1554 1555 1556
		} else {
			spin_lock(&BTRFS_I(inode)->lock);
			BTRFS_I(inode)->outstanding_extents++;
			spin_unlock(&BTRFS_I(inode)->lock);
		}
1557

1558 1559
		__percpu_counter_add(&root->fs_info->delalloc_bytes, len,
				     root->fs_info->delalloc_batch);
1560
		spin_lock(&BTRFS_I(inode)->lock);
1561
		BTRFS_I(inode)->delalloc_bytes += len;
1562 1563 1564 1565 1566 1567 1568 1569 1570 1571
		if (do_list && !test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
					 &BTRFS_I(inode)->runtime_flags)) {
			spin_lock(&root->fs_info->delalloc_lock);
			if (list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
				list_add_tail(&BTRFS_I(inode)->delalloc_inodes,
					      &root->fs_info->delalloc_inodes);
				set_bit(BTRFS_INODE_IN_DELALLOC_LIST,
					&BTRFS_I(inode)->runtime_flags);
			}
			spin_unlock(&root->fs_info->delalloc_lock);
1572
		}
1573
		spin_unlock(&BTRFS_I(inode)->lock);
1574 1575 1576
	}
}

C
Chris Mason 已提交
1577 1578 1579
/*
 * extent_io.c clear_bit_hook, see set_bit_hook for why
 */
1580
static void btrfs_clear_bit_hook(struct inode *inode,
1581 1582
				 struct extent_state *state,
				 unsigned long *bits)
1583
{
1584 1585
	/*
	 * set_bit and clear bit hooks normally require _irqsave/restore
1586
	 * but in this case, we are only testing for the DELALLOC
1587 1588
	 * bit, which is only set or cleared with irqs on
	 */
1589
	if ((state->state & EXTENT_DELALLOC) && (*bits & EXTENT_DELALLOC)) {
1590
		struct btrfs_root *root = BTRFS_I(inode)->root;
1591
		u64 len = state->end + 1 - state->start;
1592
		bool do_list = !btrfs_is_free_space_inode(inode);
1593

1594
		if (*bits & EXTENT_FIRST_DELALLOC) {
1595
			*bits &= ~EXTENT_FIRST_DELALLOC;
1596 1597 1598 1599 1600
		} else if (!(*bits & EXTENT_DO_ACCOUNTING)) {
			spin_lock(&BTRFS_I(inode)->lock);
			BTRFS_I(inode)->outstanding_extents--;
			spin_unlock(&BTRFS_I(inode)->lock);
		}
1601 1602 1603 1604

		if (*bits & EXTENT_DO_ACCOUNTING)
			btrfs_delalloc_release_metadata(inode, len);

J
Josef Bacik 已提交
1605 1606
		if (root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID
		    && do_list)
1607
			btrfs_free_reserved_data_space(inode, len);
J
Josef Bacik 已提交
1608

1609 1610
		__percpu_counter_add(&root->fs_info->delalloc_bytes, -len,
				     root->fs_info->delalloc_batch);
1611
		spin_lock(&BTRFS_I(inode)->lock);
1612
		BTRFS_I(inode)->delalloc_bytes -= len;
J
Josef Bacik 已提交
1613
		if (do_list && BTRFS_I(inode)->delalloc_bytes == 0 &&
1614 1615 1616 1617 1618 1619 1620 1621 1622
		    test_bit(BTRFS_INODE_IN_DELALLOC_LIST,
			     &BTRFS_I(inode)->runtime_flags)) {
			spin_lock(&root->fs_info->delalloc_lock);
			if (!list_empty(&BTRFS_I(inode)->delalloc_inodes)) {
				list_del_init(&BTRFS_I(inode)->delalloc_inodes);
				clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
					  &BTRFS_I(inode)->runtime_flags);
			}
			spin_unlock(&root->fs_info->delalloc_lock);
1623
		}
1624
		spin_unlock(&BTRFS_I(inode)->lock);
1625 1626 1627
	}
}

C
Chris Mason 已提交
1628 1629 1630 1631
/*
 * extent_io.c merge_bio_hook, this must check the chunk tree to make sure
 * we don't create bios that span stripes or chunks
 */
1632
int btrfs_merge_bio_hook(int rw, struct page *page, unsigned long offset,
C
Chris Mason 已提交
1633 1634
			 size_t size, struct bio *bio,
			 unsigned long bio_flags)
1635 1636
{
	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
1637
	u64 logical = (u64)bio->bi_sector << 9;
1638 1639 1640 1641
	u64 length = 0;
	u64 map_length;
	int ret;

1642 1643 1644
	if (bio_flags & EXTENT_BIO_COMPRESSED)
		return 0;

1645
	length = bio->bi_size;
1646
	map_length = length;
1647
	ret = btrfs_map_block(root->fs_info, rw, logical,
1648
			      &map_length, NULL, 0);
1649
	/* Will always return 0 with map_multi == NULL */
1650
	BUG_ON(ret < 0);
C
Chris Mason 已提交
1651
	if (map_length < length + size)
1652
		return 1;
1653
	return 0;
1654 1655
}

C
Chris Mason 已提交
1656 1657 1658 1659 1660 1661 1662 1663
/*
 * in order to insert checksums into the metadata in large chunks,
 * we wait until bio submission time.   All the pages in the bio are
 * checksummed and sums are attached onto the ordered extent record.
 *
 * At IO completion time the cums attached on the ordered extent record
 * are inserted into the btree
 */
C
Chris Mason 已提交
1664 1665
static int __btrfs_submit_bio_start(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
1666 1667
				    unsigned long bio_flags,
				    u64 bio_offset)
1668 1669 1670
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1671

1672
	ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
1673
	BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
1674 1675
	return 0;
}
1676

C
Chris Mason 已提交
1677 1678 1679 1680 1681 1682 1683 1684
/*
 * in order to insert checksums into the metadata in large chunks,
 * we wait until bio submission time.   All the pages in the bio are
 * checksummed and sums are attached onto the ordered extent record.
 *
 * At IO completion time the cums attached on the ordered extent record
 * are inserted into the btree
 */
1685
static int __btrfs_submit_bio_done(struct inode *inode, int rw, struct bio *bio,
1686 1687
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
C
Chris Mason 已提交
1688 1689
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
1690 1691 1692 1693 1694 1695
	int ret;

	ret = btrfs_map_bio(root, rw, bio, mirror_num, 1);
	if (ret)
		bio_endio(bio, ret);
	return ret;
1696 1697
}

C
Chris Mason 已提交
1698
/*
1699 1700
 * extent_io.c submission hook. This does the right thing for csum calculation
 * on write, or reading the csums from the tree before a read
C
Chris Mason 已提交
1701
 */
1702
static int btrfs_submit_bio_hook(struct inode *inode, int rw, struct bio *bio,
1703 1704
			  int mirror_num, unsigned long bio_flags,
			  u64 bio_offset)
1705 1706 1707
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret = 0;
1708
	int skip_sum;
1709
	int metadata = 0;
1710
	int async = !atomic_read(&BTRFS_I(inode)->sync_writers);
1711

1712
	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;
1713

1714
	if (btrfs_is_free_space_inode(inode))
1715 1716
		metadata = 2;

1717
	if (!(rw & REQ_WRITE)) {
1718 1719
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, metadata);
		if (ret)
1720
			goto out;
1721

1722
		if (bio_flags & EXTENT_BIO_COMPRESSED) {
1723 1724 1725 1726
			ret = btrfs_submit_compressed_read(inode, bio,
							   mirror_num,
							   bio_flags);
			goto out;
1727 1728 1729
		} else if (!skip_sum) {
			ret = btrfs_lookup_bio_sums(root, inode, bio, NULL);
			if (ret)
1730
				goto out;
1731
		}
1732
		goto mapit;
1733
	} else if (async && !skip_sum) {
1734 1735 1736
		/* csum items have already been cloned */
		if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID)
			goto mapit;
1737
		/* we're doing a write, do the async checksumming */
1738
		ret = btrfs_wq_submit_bio(BTRFS_I(inode)->root->fs_info,
1739
				   inode, rw, bio, mirror_num,
1740 1741
				   bio_flags, bio_offset,
				   __btrfs_submit_bio_start,
C
Chris Mason 已提交
1742
				   __btrfs_submit_bio_done);
1743
		goto out;
1744 1745 1746 1747
	} else if (!skip_sum) {
		ret = btrfs_csum_one_bio(root, inode, bio, 0, 0);
		if (ret)
			goto out;
1748 1749
	}

1750
mapit:
1751 1752 1753 1754 1755 1756
	ret = btrfs_map_bio(root, rw, bio, mirror_num, 0);

out:
	if (ret < 0)
		bio_endio(bio, ret);
	return ret;
1757
}
C
Chris Mason 已提交
1758

C
Chris Mason 已提交
1759 1760 1761 1762
/*
 * given a list of ordered sums record them in the inode.  This happens
 * at IO completion time based on sums calculated at bio submission time.
 */
1763
static noinline int add_pending_csums(struct btrfs_trans_handle *trans,
1764 1765 1766 1767 1768
			     struct inode *inode, u64 file_offset,
			     struct list_head *list)
{
	struct btrfs_ordered_sum *sum;

Q
Qinghuang Feng 已提交
1769
	list_for_each_entry(sum, list, list) {
M
Miao Xie 已提交
1770
		trans->adding_csums = 1;
1771 1772
		btrfs_csum_file_blocks(trans,
		       BTRFS_I(inode)->root->fs_info->csum_root, sum);
M
Miao Xie 已提交
1773
		trans->adding_csums = 0;
1774 1775 1776 1777
	}
	return 0;
}

1778 1779
int btrfs_set_extent_delalloc(struct inode *inode, u64 start, u64 end,
			      struct extent_state **cached_state)
1780
{
J
Julia Lawall 已提交
1781
	WARN_ON((end & (PAGE_CACHE_SIZE - 1)) == 0);
1782
	return set_extent_delalloc(&BTRFS_I(inode)->io_tree, start, end,
1783
				   cached_state, GFP_NOFS);
1784 1785
}

C
Chris Mason 已提交
1786
/* see btrfs_writepage_start_hook for details on why this is required */
1787 1788 1789 1790 1791
struct btrfs_writepage_fixup {
	struct page *page;
	struct btrfs_work work;
};

1792
static void btrfs_writepage_fixup_worker(struct btrfs_work *work)
1793 1794 1795
{
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_ordered_extent *ordered;
1796
	struct extent_state *cached_state = NULL;
1797 1798 1799 1800
	struct page *page;
	struct inode *inode;
	u64 page_start;
	u64 page_end;
1801
	int ret;
1802 1803 1804

	fixup = container_of(work, struct btrfs_writepage_fixup, work);
	page = fixup->page;
C
Chris Mason 已提交
1805
again:
1806 1807 1808 1809 1810 1811 1812 1813 1814 1815
	lock_page(page);
	if (!page->mapping || !PageDirty(page) || !PageChecked(page)) {
		ClearPageChecked(page);
		goto out_page;
	}

	inode = page->mapping->host;
	page_start = page_offset(page);
	page_end = page_offset(page) + PAGE_CACHE_SIZE - 1;

1816
	lock_extent_bits(&BTRFS_I(inode)->io_tree, page_start, page_end, 0,
1817
			 &cached_state);
C
Chris Mason 已提交
1818 1819

	/* already ordered? We're done */
1820
	if (PagePrivate2(page))
1821
		goto out;
C
Chris Mason 已提交
1822 1823 1824

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
1825 1826
		unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start,
				     page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
1827 1828
		unlock_page(page);
		btrfs_start_ordered_extent(inode, ordered, 1);
1829
		btrfs_put_ordered_extent(ordered);
C
Chris Mason 已提交
1830 1831
		goto again;
	}
1832

1833 1834 1835 1836 1837 1838 1839 1840
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
	if (ret) {
		mapping_set_error(page->mapping, ret);
		end_extent_writepage(page, ret, page_start, page_end);
		ClearPageChecked(page);
		goto out;
	 }

1841
	btrfs_set_extent_delalloc(inode, page_start, page_end, &cached_state);
1842
	ClearPageChecked(page);
1843
	set_page_dirty(page);
1844
out:
1845 1846
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, page_start, page_end,
			     &cached_state, GFP_NOFS);
1847 1848 1849
out_page:
	unlock_page(page);
	page_cache_release(page);
1850
	kfree(fixup);
1851 1852 1853 1854 1855 1856 1857 1858
}

/*
 * There are a few paths in the higher layers of the kernel that directly
 * set the page dirty bit without asking the filesystem if it is a
 * good idea.  This causes problems because we want to make sure COW
 * properly happens and the data=ordered rules are followed.
 *
C
Chris Mason 已提交
1859
 * In our case any range that doesn't have the ORDERED bit set
1860 1861 1862 1863
 * hasn't been properly setup for IO.  We kick off an async process
 * to fix it up.  The async helper will wait for ordered extents, set
 * the delalloc bit and make it safe to write the page.
 */
1864
static int btrfs_writepage_start_hook(struct page *page, u64 start, u64 end)
1865 1866 1867 1868 1869
{
	struct inode *inode = page->mapping->host;
	struct btrfs_writepage_fixup *fixup;
	struct btrfs_root *root = BTRFS_I(inode)->root;

1870 1871
	/* this page is properly in the ordered list */
	if (TestClearPagePrivate2(page))
1872 1873 1874 1875 1876 1877 1878 1879
		return 0;

	if (PageChecked(page))
		return -EAGAIN;

	fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
	if (!fixup)
		return -EAGAIN;
1880

1881 1882 1883 1884 1885
	SetPageChecked(page);
	page_cache_get(page);
	fixup->work.func = btrfs_writepage_fixup_worker;
	fixup->page = page;
	btrfs_queue_worker(&root->fs_info->fixup_workers, &fixup->work);
1886
	return -EBUSY;
1887 1888
}

Y
Yan Zheng 已提交
1889 1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901 1902 1903
static int insert_reserved_file_extent(struct btrfs_trans_handle *trans,
				       struct inode *inode, u64 file_pos,
				       u64 disk_bytenr, u64 disk_num_bytes,
				       u64 num_bytes, u64 ram_bytes,
				       u8 compression, u8 encryption,
				       u16 other_encoding, int extent_type)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key ins;
	int ret;

	path = btrfs_alloc_path();
1904 1905
	if (!path)
		return -ENOMEM;
Y
Yan Zheng 已提交
1906

1907
	path->leave_spinning = 1;
C
Chris Mason 已提交
1908 1909 1910 1911 1912 1913 1914 1915 1916 1917

	/*
	 * we may be replacing one extent in the tree with another.
	 * The new extent is pinned in the extent map, and we don't want
	 * to drop it from the cache until it is completely in the btree.
	 *
	 * So, tell btrfs_drop_extents to leave this extent in the cache.
	 * the caller is expected to unpin it and allow it to be merged
	 * with the others.
	 */
J
Josef Bacik 已提交
1918
	ret = btrfs_drop_extents(trans, root, inode, file_pos,
1919
				 file_pos + num_bytes, 0);
1920 1921
	if (ret)
		goto out;
Y
Yan Zheng 已提交
1922

L
Li Zefan 已提交
1923
	ins.objectid = btrfs_ino(inode);
Y
Yan Zheng 已提交
1924 1925 1926
	ins.offset = file_pos;
	ins.type = BTRFS_EXTENT_DATA_KEY;
	ret = btrfs_insert_empty_item(trans, root, path, &ins, sizeof(*fi));
1927 1928
	if (ret)
		goto out;
Y
Yan Zheng 已提交
1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
	leaf = path->nodes[0];
	fi = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, fi, trans->transid);
	btrfs_set_file_extent_type(leaf, fi, extent_type);
	btrfs_set_file_extent_disk_bytenr(leaf, fi, disk_bytenr);
	btrfs_set_file_extent_disk_num_bytes(leaf, fi, disk_num_bytes);
	btrfs_set_file_extent_offset(leaf, fi, 0);
	btrfs_set_file_extent_num_bytes(leaf, fi, num_bytes);
	btrfs_set_file_extent_ram_bytes(leaf, fi, ram_bytes);
	btrfs_set_file_extent_compression(leaf, fi, compression);
	btrfs_set_file_extent_encryption(leaf, fi, encryption);
	btrfs_set_file_extent_other_encoding(leaf, fi, other_encoding);
1942

Y
Yan Zheng 已提交
1943
	btrfs_mark_buffer_dirty(leaf);
1944
	btrfs_release_path(path);
Y
Yan Zheng 已提交
1945 1946 1947 1948 1949 1950

	inode_add_bytes(inode, num_bytes);

	ins.objectid = disk_bytenr;
	ins.offset = disk_num_bytes;
	ins.type = BTRFS_EXTENT_ITEM_KEY;
1951 1952
	ret = btrfs_alloc_reserved_file_extent(trans, root,
					root->root_key.objectid,
L
Li Zefan 已提交
1953
					btrfs_ino(inode), file_pos, &ins);
1954
out:
Y
Yan Zheng 已提交
1955
	btrfs_free_path(path);
1956

1957
	return ret;
Y
Yan Zheng 已提交
1958 1959
}

L
Liu Bo 已提交
1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254 2255 2256 2257 2258 2259 2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286 2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340
/* snapshot-aware defrag */
struct sa_defrag_extent_backref {
	struct rb_node node;
	struct old_sa_defrag_extent *old;
	u64 root_id;
	u64 inum;
	u64 file_pos;
	u64 extent_offset;
	u64 num_bytes;
	u64 generation;
};

struct old_sa_defrag_extent {
	struct list_head list;
	struct new_sa_defrag_extent *new;

	u64 extent_offset;
	u64 bytenr;
	u64 offset;
	u64 len;
	int count;
};

struct new_sa_defrag_extent {
	struct rb_root root;
	struct list_head head;
	struct btrfs_path *path;
	struct inode *inode;
	u64 file_pos;
	u64 len;
	u64 bytenr;
	u64 disk_len;
	u8 compress_type;
};

static int backref_comp(struct sa_defrag_extent_backref *b1,
			struct sa_defrag_extent_backref *b2)
{
	if (b1->root_id < b2->root_id)
		return -1;
	else if (b1->root_id > b2->root_id)
		return 1;

	if (b1->inum < b2->inum)
		return -1;
	else if (b1->inum > b2->inum)
		return 1;

	if (b1->file_pos < b2->file_pos)
		return -1;
	else if (b1->file_pos > b2->file_pos)
		return 1;

	/*
	 * [------------------------------] ===> (a range of space)
	 *     |<--->|   |<---->| =============> (fs/file tree A)
	 * |<---------------------------->| ===> (fs/file tree B)
	 *
	 * A range of space can refer to two file extents in one tree while
	 * refer to only one file extent in another tree.
	 *
	 * So we may process a disk offset more than one time(two extents in A)
	 * and locate at the same extent(one extent in B), then insert two same
	 * backrefs(both refer to the extent in B).
	 */
	return 0;
}

static void backref_insert(struct rb_root *root,
			   struct sa_defrag_extent_backref *backref)
{
	struct rb_node **p = &root->rb_node;
	struct rb_node *parent = NULL;
	struct sa_defrag_extent_backref *entry;
	int ret;

	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct sa_defrag_extent_backref, node);

		ret = backref_comp(backref, entry);
		if (ret < 0)
			p = &(*p)->rb_left;
		else
			p = &(*p)->rb_right;
	}

	rb_link_node(&backref->node, parent, p);
	rb_insert_color(&backref->node, root);
}

/*
 * Note the backref might has changed, and in this case we just return 0.
 */
static noinline int record_one_backref(u64 inum, u64 offset, u64 root_id,
				       void *ctx)
{
	struct btrfs_file_extent_item *extent;
	struct btrfs_fs_info *fs_info;
	struct old_sa_defrag_extent *old = ctx;
	struct new_sa_defrag_extent *new = old->new;
	struct btrfs_path *path = new->path;
	struct btrfs_key key;
	struct btrfs_root *root;
	struct sa_defrag_extent_backref *backref;
	struct extent_buffer *leaf;
	struct inode *inode = new->inode;
	int slot;
	int ret;
	u64 extent_offset;
	u64 num_bytes;

	if (BTRFS_I(inode)->root->root_key.objectid == root_id &&
	    inum == btrfs_ino(inode))
		return 0;

	key.objectid = root_id;
	key.type = BTRFS_ROOT_ITEM_KEY;
	key.offset = (u64)-1;

	fs_info = BTRFS_I(inode)->root->fs_info;
	root = btrfs_read_fs_root_no_name(fs_info, &key);
	if (IS_ERR(root)) {
		if (PTR_ERR(root) == -ENOENT)
			return 0;
		WARN_ON(1);
		pr_debug("inum=%llu, offset=%llu, root_id=%llu\n",
			 inum, offset, root_id);
		return PTR_ERR(root);
	}

	key.objectid = inum;
	key.type = BTRFS_EXTENT_DATA_KEY;
	if (offset > (u64)-1 << 32)
		key.offset = 0;
	else
		key.offset = offset;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0) {
		WARN_ON(1);
		return ret;
	}

	while (1) {
		cond_resched();

		leaf = path->nodes[0];
		slot = path->slots[0];

		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0) {
				goto out;
			} else if (ret > 0) {
				ret = 0;
				goto out;
			}
			continue;
		}

		path->slots[0]++;

		btrfs_item_key_to_cpu(leaf, &key, slot);

		if (key.objectid > inum)
			goto out;

		if (key.objectid < inum || key.type != BTRFS_EXTENT_DATA_KEY)
			continue;

		extent = btrfs_item_ptr(leaf, slot,
					struct btrfs_file_extent_item);

		if (btrfs_file_extent_disk_bytenr(leaf, extent) != old->bytenr)
			continue;

		extent_offset = btrfs_file_extent_offset(leaf, extent);
		if (key.offset - extent_offset != offset)
			continue;

		num_bytes = btrfs_file_extent_num_bytes(leaf, extent);
		if (extent_offset >= old->extent_offset + old->offset +
		    old->len || extent_offset + num_bytes <=
		    old->extent_offset + old->offset)
			continue;

		break;
	}

	backref = kmalloc(sizeof(*backref), GFP_NOFS);
	if (!backref) {
		ret = -ENOENT;
		goto out;
	}

	backref->root_id = root_id;
	backref->inum = inum;
	backref->file_pos = offset + extent_offset;
	backref->num_bytes = num_bytes;
	backref->extent_offset = extent_offset;
	backref->generation = btrfs_file_extent_generation(leaf, extent);
	backref->old = old;
	backref_insert(&new->root, backref);
	old->count++;
out:
	btrfs_release_path(path);
	WARN_ON(ret);
	return ret;
}

static noinline bool record_extent_backrefs(struct btrfs_path *path,
				   struct new_sa_defrag_extent *new)
{
	struct btrfs_fs_info *fs_info = BTRFS_I(new->inode)->root->fs_info;
	struct old_sa_defrag_extent *old, *tmp;
	int ret;

	new->path = path;

	list_for_each_entry_safe(old, tmp, &new->head, list) {
		ret = iterate_inodes_from_logical(old->bytenr, fs_info,
						  path, record_one_backref,
						  old);
		BUG_ON(ret < 0 && ret != -ENOENT);

		/* no backref to be processed for this extent */
		if (!old->count) {
			list_del(&old->list);
			kfree(old);
		}
	}

	if (list_empty(&new->head))
		return false;

	return true;
}

static int relink_is_mergable(struct extent_buffer *leaf,
			      struct btrfs_file_extent_item *fi,
			      u64 disk_bytenr)
{
	if (btrfs_file_extent_disk_bytenr(leaf, fi) != disk_bytenr)
		return 0;

	if (btrfs_file_extent_type(leaf, fi) != BTRFS_FILE_EXTENT_REG)
		return 0;

	if (btrfs_file_extent_compression(leaf, fi) ||
	    btrfs_file_extent_encryption(leaf, fi) ||
	    btrfs_file_extent_other_encoding(leaf, fi))
		return 0;

	return 1;
}

/*
 * Note the backref might has changed, and in this case we just return 0.
 */
static noinline int relink_extent_backref(struct btrfs_path *path,
				 struct sa_defrag_extent_backref *prev,
				 struct sa_defrag_extent_backref *backref)
{
	struct btrfs_file_extent_item *extent;
	struct btrfs_file_extent_item *item;
	struct btrfs_ordered_extent *ordered;
	struct btrfs_trans_handle *trans;
	struct btrfs_fs_info *fs_info;
	struct btrfs_root *root;
	struct btrfs_key key;
	struct extent_buffer *leaf;
	struct old_sa_defrag_extent *old = backref->old;
	struct new_sa_defrag_extent *new = old->new;
	struct inode *src_inode = new->inode;
	struct inode *inode;
	struct extent_state *cached = NULL;
	int ret = 0;
	u64 start;
	u64 len;
	u64 lock_start;
	u64 lock_end;
	bool merge = false;
	int index;

	if (prev && prev->root_id == backref->root_id &&
	    prev->inum == backref->inum &&
	    prev->file_pos + prev->num_bytes == backref->file_pos)
		merge = true;

	/* step 1: get root */
	key.objectid = backref->root_id;
	key.type = BTRFS_ROOT_ITEM_KEY;
	key.offset = (u64)-1;

	fs_info = BTRFS_I(src_inode)->root->fs_info;
	index = srcu_read_lock(&fs_info->subvol_srcu);

	root = btrfs_read_fs_root_no_name(fs_info, &key);
	if (IS_ERR(root)) {
		srcu_read_unlock(&fs_info->subvol_srcu, index);
		if (PTR_ERR(root) == -ENOENT)
			return 0;
		return PTR_ERR(root);
	}
	if (btrfs_root_refs(&root->root_item) == 0) {
		srcu_read_unlock(&fs_info->subvol_srcu, index);
		/* parse ENOENT to 0 */
		return 0;
	}

	/* step 2: get inode */
	key.objectid = backref->inum;
	key.type = BTRFS_INODE_ITEM_KEY;
	key.offset = 0;

	inode = btrfs_iget(fs_info->sb, &key, root, NULL);
	if (IS_ERR(inode)) {
		srcu_read_unlock(&fs_info->subvol_srcu, index);
		return 0;
	}

	srcu_read_unlock(&fs_info->subvol_srcu, index);

	/* step 3: relink backref */
	lock_start = backref->file_pos;
	lock_end = backref->file_pos + backref->num_bytes - 1;
	lock_extent_bits(&BTRFS_I(inode)->io_tree, lock_start, lock_end,
			 0, &cached);

	ordered = btrfs_lookup_first_ordered_extent(inode, lock_end);
	if (ordered) {
		btrfs_put_ordered_extent(ordered);
		goto out_unlock;
	}

	trans = btrfs_join_transaction(root);
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
		goto out_unlock;
	}

	key.objectid = backref->inum;
	key.type = BTRFS_EXTENT_DATA_KEY;
	key.offset = backref->file_pos;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0) {
		goto out_free_path;
	} else if (ret > 0) {
		ret = 0;
		goto out_free_path;
	}

	extent = btrfs_item_ptr(path->nodes[0], path->slots[0],
				struct btrfs_file_extent_item);

	if (btrfs_file_extent_generation(path->nodes[0], extent) !=
	    backref->generation)
		goto out_free_path;

	btrfs_release_path(path);

	start = backref->file_pos;
	if (backref->extent_offset < old->extent_offset + old->offset)
		start += old->extent_offset + old->offset -
			 backref->extent_offset;

	len = min(backref->extent_offset + backref->num_bytes,
		  old->extent_offset + old->offset + old->len);
	len -= max(backref->extent_offset, old->extent_offset + old->offset);

	ret = btrfs_drop_extents(trans, root, inode, start,
				 start + len, 1);
	if (ret)
		goto out_free_path;
again:
	key.objectid = btrfs_ino(inode);
	key.type = BTRFS_EXTENT_DATA_KEY;
	key.offset = start;

2341
	path->leave_spinning = 1;
L
Liu Bo 已提交
2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397
	if (merge) {
		struct btrfs_file_extent_item *fi;
		u64 extent_len;
		struct btrfs_key found_key;

		ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
		if (ret < 0)
			goto out_free_path;

		path->slots[0]--;
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

		fi = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_file_extent_item);
		extent_len = btrfs_file_extent_num_bytes(leaf, fi);

		if (relink_is_mergable(leaf, fi, new->bytenr) &&
		    extent_len + found_key.offset == start) {
			btrfs_set_file_extent_num_bytes(leaf, fi,
							extent_len + len);
			btrfs_mark_buffer_dirty(leaf);
			inode_add_bytes(inode, len);

			ret = 1;
			goto out_free_path;
		} else {
			merge = false;
			btrfs_release_path(path);
			goto again;
		}
	}

	ret = btrfs_insert_empty_item(trans, root, path, &key,
					sizeof(*extent));
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_free_path;
	}

	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
				struct btrfs_file_extent_item);
	btrfs_set_file_extent_disk_bytenr(leaf, item, new->bytenr);
	btrfs_set_file_extent_disk_num_bytes(leaf, item, new->disk_len);
	btrfs_set_file_extent_offset(leaf, item, start - new->file_pos);
	btrfs_set_file_extent_num_bytes(leaf, item, len);
	btrfs_set_file_extent_ram_bytes(leaf, item, new->len);
	btrfs_set_file_extent_generation(leaf, item, trans->transid);
	btrfs_set_file_extent_type(leaf, item, BTRFS_FILE_EXTENT_REG);
	btrfs_set_file_extent_compression(leaf, item, new->compress_type);
	btrfs_set_file_extent_encryption(leaf, item, 0);
	btrfs_set_file_extent_other_encoding(leaf, item, 0);

	btrfs_mark_buffer_dirty(leaf);
	inode_add_bytes(inode, len);
2398
	btrfs_release_path(path);
L
Liu Bo 已提交
2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410 2411

	ret = btrfs_inc_extent_ref(trans, root, new->bytenr,
			new->disk_len, 0,
			backref->root_id, backref->inum,
			new->file_pos, 0);	/* start - extent_offset */
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_free_path;
	}

	ret = 1;
out_free_path:
	btrfs_release_path(path);
2412
	path->leave_spinning = 0;
L
Liu Bo 已提交
2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423 2424 2425 2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522 2523 2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560 2561 2562 2563 2564 2565 2566 2567 2568 2569 2570 2571 2572 2573 2574 2575 2576 2577 2578 2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596
	btrfs_end_transaction(trans, root);
out_unlock:
	unlock_extent_cached(&BTRFS_I(inode)->io_tree, lock_start, lock_end,
			     &cached, GFP_NOFS);
	iput(inode);
	return ret;
}

static void relink_file_extents(struct new_sa_defrag_extent *new)
{
	struct btrfs_path *path;
	struct old_sa_defrag_extent *old, *tmp;
	struct sa_defrag_extent_backref *backref;
	struct sa_defrag_extent_backref *prev = NULL;
	struct inode *inode;
	struct btrfs_root *root;
	struct rb_node *node;
	int ret;

	inode = new->inode;
	root = BTRFS_I(inode)->root;

	path = btrfs_alloc_path();
	if (!path)
		return;

	if (!record_extent_backrefs(path, new)) {
		btrfs_free_path(path);
		goto out;
	}
	btrfs_release_path(path);

	while (1) {
		node = rb_first(&new->root);
		if (!node)
			break;
		rb_erase(node, &new->root);

		backref = rb_entry(node, struct sa_defrag_extent_backref, node);

		ret = relink_extent_backref(path, prev, backref);
		WARN_ON(ret < 0);

		kfree(prev);

		if (ret == 1)
			prev = backref;
		else
			prev = NULL;
		cond_resched();
	}
	kfree(prev);

	btrfs_free_path(path);

	list_for_each_entry_safe(old, tmp, &new->head, list) {
		list_del(&old->list);
		kfree(old);
	}
out:
	atomic_dec(&root->fs_info->defrag_running);
	wake_up(&root->fs_info->transaction_wait);

	kfree(new);
}

static struct new_sa_defrag_extent *
record_old_file_extents(struct inode *inode,
			struct btrfs_ordered_extent *ordered)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_path *path;
	struct btrfs_key key;
	struct old_sa_defrag_extent *old, *tmp;
	struct new_sa_defrag_extent *new;
	int ret;

	new = kmalloc(sizeof(*new), GFP_NOFS);
	if (!new)
		return NULL;

	new->inode = inode;
	new->file_pos = ordered->file_offset;
	new->len = ordered->len;
	new->bytenr = ordered->start;
	new->disk_len = ordered->disk_len;
	new->compress_type = ordered->compress_type;
	new->root = RB_ROOT;
	INIT_LIST_HEAD(&new->head);

	path = btrfs_alloc_path();
	if (!path)
		goto out_kfree;

	key.objectid = btrfs_ino(inode);
	key.type = BTRFS_EXTENT_DATA_KEY;
	key.offset = new->file_pos;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto out_free_path;
	if (ret > 0 && path->slots[0] > 0)
		path->slots[0]--;

	/* find out all the old extents for the file range */
	while (1) {
		struct btrfs_file_extent_item *extent;
		struct extent_buffer *l;
		int slot;
		u64 num_bytes;
		u64 offset;
		u64 end;
		u64 disk_bytenr;
		u64 extent_offset;

		l = path->nodes[0];
		slot = path->slots[0];

		if (slot >= btrfs_header_nritems(l)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto out_free_list;
			else if (ret > 0)
				break;
			continue;
		}

		btrfs_item_key_to_cpu(l, &key, slot);

		if (key.objectid != btrfs_ino(inode))
			break;
		if (key.type != BTRFS_EXTENT_DATA_KEY)
			break;
		if (key.offset >= new->file_pos + new->len)
			break;

		extent = btrfs_item_ptr(l, slot, struct btrfs_file_extent_item);

		num_bytes = btrfs_file_extent_num_bytes(l, extent);
		if (key.offset + num_bytes < new->file_pos)
			goto next;

		disk_bytenr = btrfs_file_extent_disk_bytenr(l, extent);
		if (!disk_bytenr)
			goto next;

		extent_offset = btrfs_file_extent_offset(l, extent);

		old = kmalloc(sizeof(*old), GFP_NOFS);
		if (!old)
			goto out_free_list;

		offset = max(new->file_pos, key.offset);
		end = min(new->file_pos + new->len, key.offset + num_bytes);

		old->bytenr = disk_bytenr;
		old->extent_offset = extent_offset;
		old->offset = offset - key.offset;
		old->len = end - offset;
		old->new = new;
		old->count = 0;
		list_add_tail(&old->list, &new->head);
next:
		path->slots[0]++;
		cond_resched();
	}

	btrfs_free_path(path);
	atomic_inc(&root->fs_info->defrag_running);

	return new;

out_free_list:
	list_for_each_entry_safe(old, tmp, &new->head, list) {
		list_del(&old->list);
		kfree(old);
	}
out_free_path:
	btrfs_free_path(path);
out_kfree:
	kfree(new);
	return NULL;
}

2597 2598 2599 2600 2601 2602
/*
 * helper function for btrfs_finish_ordered_io, this
 * just reads in some of the csum leaves to prime them into ram
 * before we start the transaction.  It limits the amount of btree
 * reads required while inside the transaction.
 */
C
Chris Mason 已提交
2603 2604 2605 2606
/* as ordered data IO finishes, this gets called so we can finish
 * an ordered extent if the range of bytes in the file it covers are
 * fully written.
 */
2607
static int btrfs_finish_ordered_io(struct btrfs_ordered_extent *ordered_extent)
2608
{
2609
	struct inode *inode = ordered_extent->inode;
2610
	struct btrfs_root *root = BTRFS_I(inode)->root;
2611
	struct btrfs_trans_handle *trans = NULL;
2612
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2613
	struct extent_state *cached_state = NULL;
L
Liu Bo 已提交
2614
	struct new_sa_defrag_extent *new = NULL;
2615
	int compress_type = 0;
2616
	int ret;
2617
	bool nolock;
2618

2619
	nolock = btrfs_is_free_space_inode(inode);
J
Josef Bacik 已提交
2620

2621 2622 2623 2624 2625
	if (test_bit(BTRFS_ORDERED_IOERR, &ordered_extent->flags)) {
		ret = -EIO;
		goto out;
	}

2626
	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags)) {
2627
		BUG_ON(!list_empty(&ordered_extent->list)); /* Logic error */
2628 2629 2630 2631 2632 2633 2634 2635 2636
		btrfs_ordered_update_i_size(inode, 0, ordered_extent);
		if (nolock)
			trans = btrfs_join_transaction_nolock(root);
		else
			trans = btrfs_join_transaction(root);
		if (IS_ERR(trans)) {
			ret = PTR_ERR(trans);
			trans = NULL;
			goto out;
2637
		}
2638 2639 2640 2641
		trans->block_rsv = &root->fs_info->delalloc_block_rsv;
		ret = btrfs_update_inode_fallback(trans, root, inode);
		if (ret) /* -ENOMEM or corruption */
			btrfs_abort_transaction(trans, root, ret);
2642 2643
		goto out;
	}
2644

2645 2646
	lock_extent_bits(io_tree, ordered_extent->file_offset,
			 ordered_extent->file_offset + ordered_extent->len - 1,
2647
			 0, &cached_state);
2648

L
Liu Bo 已提交
2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662
	ret = test_range_bit(io_tree, ordered_extent->file_offset,
			ordered_extent->file_offset + ordered_extent->len - 1,
			EXTENT_DEFRAG, 1, cached_state);
	if (ret) {
		u64 last_snapshot = btrfs_root_last_snapshot(&root->root_item);
		if (last_snapshot >= BTRFS_I(inode)->generation)
			/* the inode is shared */
			new = record_old_file_extents(inode, ordered_extent);

		clear_extent_bit(io_tree, ordered_extent->file_offset,
			ordered_extent->file_offset + ordered_extent->len - 1,
			EXTENT_DEFRAG, 0, 0, &cached_state, GFP_NOFS);
	}

J
Josef Bacik 已提交
2663
	if (nolock)
2664
		trans = btrfs_join_transaction_nolock(root);
J
Josef Bacik 已提交
2665
	else
2666
		trans = btrfs_join_transaction(root);
2667 2668 2669 2670 2671
	if (IS_ERR(trans)) {
		ret = PTR_ERR(trans);
		trans = NULL;
		goto out_unlock;
	}
2672
	trans->block_rsv = &root->fs_info->delalloc_block_rsv;
2673

C
Chris Mason 已提交
2674
	if (test_bit(BTRFS_ORDERED_COMPRESSED, &ordered_extent->flags))
2675
		compress_type = ordered_extent->compress_type;
Y
Yan Zheng 已提交
2676
	if (test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags)) {
2677
		BUG_ON(compress_type);
Y
Yan, Zheng 已提交
2678
		ret = btrfs_mark_extent_written(trans, inode,
Y
Yan Zheng 已提交
2679 2680 2681 2682
						ordered_extent->file_offset,
						ordered_extent->file_offset +
						ordered_extent->len);
	} else {
2683
		BUG_ON(root == root->fs_info->tree_root);
Y
Yan Zheng 已提交
2684 2685 2686 2687 2688 2689
		ret = insert_reserved_file_extent(trans, inode,
						ordered_extent->file_offset,
						ordered_extent->start,
						ordered_extent->disk_len,
						ordered_extent->len,
						ordered_extent->len,
2690
						compress_type, 0, 0,
Y
Yan Zheng 已提交
2691 2692
						BTRFS_FILE_EXTENT_REG);
	}
J
Josef Bacik 已提交
2693 2694 2695
	unpin_extent_cache(&BTRFS_I(inode)->extent_tree,
			   ordered_extent->file_offset, ordered_extent->len,
			   trans->transid);
2696 2697
	if (ret < 0) {
		btrfs_abort_transaction(trans, root, ret);
2698
		goto out_unlock;
2699
	}
2700

2701 2702 2703
	add_pending_csums(trans, inode, ordered_extent->file_offset,
			  &ordered_extent->list);

2704 2705 2706 2707 2708
	btrfs_ordered_update_i_size(inode, 0, ordered_extent);
	ret = btrfs_update_inode_fallback(trans, root, inode);
	if (ret) { /* -ENOMEM or corruption */
		btrfs_abort_transaction(trans, root, ret);
		goto out_unlock;
2709 2710
	}
	ret = 0;
2711 2712 2713 2714
out_unlock:
	unlock_extent_cached(io_tree, ordered_extent->file_offset,
			     ordered_extent->file_offset +
			     ordered_extent->len - 1, &cached_state, GFP_NOFS);
2715
out:
2716
	if (root != root->fs_info->tree_root)
J
Josef Bacik 已提交
2717
		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
2718 2719
	if (trans)
		btrfs_end_transaction(trans, root);
J
Josef Bacik 已提交
2720

2721
	if (ret) {
2722 2723 2724 2725
		clear_extent_uptodate(io_tree, ordered_extent->file_offset,
				      ordered_extent->file_offset +
				      ordered_extent->len - 1, NULL, GFP_NOFS);

2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737
		/*
		 * If the ordered extent had an IOERR or something else went
		 * wrong we need to return the space for this ordered extent
		 * back to the allocator.
		 */
		if (!test_bit(BTRFS_ORDERED_NOCOW, &ordered_extent->flags) &&
		    !test_bit(BTRFS_ORDERED_PREALLOC, &ordered_extent->flags))
			btrfs_free_reserved_extent(root, ordered_extent->start,
						   ordered_extent->disk_len);
	}


2738
	/*
2739 2740
	 * This needs to be done to make sure anybody waiting knows we are done
	 * updating everything for this ordered extent.
2741 2742 2743
	 */
	btrfs_remove_ordered_extent(inode, ordered_extent);

L
Liu Bo 已提交
2744 2745 2746 2747
	/* for snapshot-aware defrag */
	if (new)
		relink_file_extents(new);

2748 2749 2750 2751 2752
	/* once for us */
	btrfs_put_ordered_extent(ordered_extent);
	/* once for the tree */
	btrfs_put_ordered_extent(ordered_extent);

2753 2754 2755 2756 2757 2758 2759 2760
	return ret;
}

static void finish_ordered_fn(struct btrfs_work *work)
{
	struct btrfs_ordered_extent *ordered_extent;
	ordered_extent = container_of(work, struct btrfs_ordered_extent, work);
	btrfs_finish_ordered_io(ordered_extent);
2761 2762
}

2763
static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
2764 2765
				struct extent_state *state, int uptodate)
{
2766 2767 2768 2769 2770
	struct inode *inode = page->mapping->host;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered_extent = NULL;
	struct btrfs_workers *workers;

2771 2772
	trace_btrfs_writepage_end_io_hook(page, start, end, uptodate);

2773
	ClearPagePrivate2(page);
2774 2775 2776 2777 2778 2779 2780
	if (!btrfs_dec_test_ordered_pending(inode, &ordered_extent, start,
					    end - start + 1, uptodate))
		return 0;

	ordered_extent->work.func = finish_ordered_fn;
	ordered_extent->work.flags = 0;

2781
	if (btrfs_is_free_space_inode(inode))
2782 2783 2784 2785 2786 2787
		workers = &root->fs_info->endio_freespace_worker;
	else
		workers = &root->fs_info->endio_write_workers;
	btrfs_queue_worker(workers, &ordered_extent->work);

	return 0;
2788 2789
}

C
Chris Mason 已提交
2790 2791
/*
 * when reads are done, we need to check csums to verify the data is correct
2792 2793
 * if there's a match, we allow the bio to finish.  If not, the code in
 * extent_io.c will try to find good copies for us.
C
Chris Mason 已提交
2794
 */
2795
static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
2796
			       struct extent_state *state, int mirror)
2797
{
M
Miao Xie 已提交
2798
	size_t offset = start - page_offset(page);
2799
	struct inode *inode = page->mapping->host;
2800
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
2801
	char *kaddr;
2802
	u64 private = ~(u32)0;
2803
	int ret;
2804 2805
	struct btrfs_root *root = BTRFS_I(inode)->root;
	u32 csum = ~(u32)0;
2806 2807
	static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
	                              DEFAULT_RATELIMIT_BURST);
2808

2809 2810 2811 2812
	if (PageChecked(page)) {
		ClearPageChecked(page);
		goto good;
	}
2813 2814

	if (BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)
2815
		goto good;
2816 2817

	if (root->root_key.objectid == BTRFS_DATA_RELOC_TREE_OBJECTID &&
2818
	    test_range_bit(io_tree, start, end, EXTENT_NODATASUM, 1, NULL)) {
2819 2820
		clear_extent_bits(io_tree, start, end, EXTENT_NODATASUM,
				  GFP_NOFS);
2821
		return 0;
2822
	}
2823

Y
Yan 已提交
2824
	if (state && state->start == start) {
2825 2826 2827 2828 2829
		private = state->private;
		ret = 0;
	} else {
		ret = get_state_private(io_tree, start, &private);
	}
2830
	kaddr = kmap_atomic(page);
C
Chris Mason 已提交
2831
	if (ret)
2832
		goto zeroit;
C
Chris Mason 已提交
2833

2834
	csum = btrfs_csum_data(kaddr + offset, csum,  end - start + 1);
2835
	btrfs_csum_final(csum, (char *)&csum);
C
Chris Mason 已提交
2836
	if (csum != private)
2837
		goto zeroit;
C
Chris Mason 已提交
2838

2839
	kunmap_atomic(kaddr);
2840
good:
2841 2842 2843
	return 0;

zeroit:
2844 2845 2846 2847 2848
	if (__ratelimit(&_rs))
		btrfs_info(root->fs_info, "csum failed ino %llu off %llu csum %u private %llu",
			(unsigned long long)btrfs_ino(page->mapping->host),
			(unsigned long long)start, csum,
			(unsigned long long)private);
2849 2850
	memset(kaddr + offset, 1, end - start + 1);
	flush_dcache_page(page);
2851
	kunmap_atomic(kaddr);
2852 2853
	if (private == 0)
		return 0;
2854
	return -EIO;
2855
}
2856

Y
Yan, Zheng 已提交
2857 2858 2859 2860 2861
struct delayed_iput {
	struct list_head list;
	struct inode *inode;
};

2862 2863
/* JDM: If this is fs-wide, why can't we add a pointer to
 * btrfs_inode instead and avoid the allocation? */
Y
Yan, Zheng 已提交
2864 2865 2866 2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897 2898 2899 2900 2901 2902 2903 2904
void btrfs_add_delayed_iput(struct inode *inode)
{
	struct btrfs_fs_info *fs_info = BTRFS_I(inode)->root->fs_info;
	struct delayed_iput *delayed;

	if (atomic_add_unless(&inode->i_count, -1, 1))
		return;

	delayed = kmalloc(sizeof(*delayed), GFP_NOFS | __GFP_NOFAIL);
	delayed->inode = inode;

	spin_lock(&fs_info->delayed_iput_lock);
	list_add_tail(&delayed->list, &fs_info->delayed_iputs);
	spin_unlock(&fs_info->delayed_iput_lock);
}

void btrfs_run_delayed_iputs(struct btrfs_root *root)
{
	LIST_HEAD(list);
	struct btrfs_fs_info *fs_info = root->fs_info;
	struct delayed_iput *delayed;
	int empty;

	spin_lock(&fs_info->delayed_iput_lock);
	empty = list_empty(&fs_info->delayed_iputs);
	spin_unlock(&fs_info->delayed_iput_lock);
	if (empty)
		return;

	spin_lock(&fs_info->delayed_iput_lock);
	list_splice_init(&fs_info->delayed_iputs, &list);
	spin_unlock(&fs_info->delayed_iput_lock);

	while (!list_empty(&list)) {
		delayed = list_entry(list.next, struct delayed_iput, list);
		list_del(&delayed->list);
		iput(delayed->inode);
		kfree(delayed);
	}
}

2905
/*
2906
 * This is called in transaction commit time. If there are no orphan
2907 2908 2909 2910 2911 2912
 * files in the subvolume, it removes orphan item and frees block_rsv
 * structure.
 */
void btrfs_orphan_commit_root(struct btrfs_trans_handle *trans,
			      struct btrfs_root *root)
{
2913
	struct btrfs_block_rsv *block_rsv;
2914 2915
	int ret;

2916
	if (atomic_read(&root->orphan_inodes) ||
2917 2918 2919
	    root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE)
		return;

2920
	spin_lock(&root->orphan_lock);
2921
	if (atomic_read(&root->orphan_inodes)) {
2922 2923 2924 2925 2926 2927 2928 2929 2930 2931 2932 2933 2934
		spin_unlock(&root->orphan_lock);
		return;
	}

	if (root->orphan_cleanup_state != ORPHAN_CLEANUP_DONE) {
		spin_unlock(&root->orphan_lock);
		return;
	}

	block_rsv = root->orphan_block_rsv;
	root->orphan_block_rsv = NULL;
	spin_unlock(&root->orphan_lock);

2935 2936 2937 2938 2939 2940 2941 2942
	if (root->orphan_item_inserted &&
	    btrfs_root_refs(&root->root_item) > 0) {
		ret = btrfs_del_orphan_item(trans, root->fs_info->tree_root,
					    root->root_key.objectid);
		BUG_ON(ret);
		root->orphan_item_inserted = 0;
	}

2943 2944 2945
	if (block_rsv) {
		WARN_ON(block_rsv->size > 0);
		btrfs_free_block_rsv(root, block_rsv);
2946 2947 2948
	}
}

2949 2950 2951
/*
 * This creates an orphan entry for the given inode in case something goes
 * wrong in the middle of an unlink/truncate.
2952 2953 2954
 *
 * NOTE: caller of this function should reserve 5 units of metadata for
 *	 this function.
2955 2956 2957 2958
 */
int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
2959 2960 2961 2962
	struct btrfs_block_rsv *block_rsv = NULL;
	int reserve = 0;
	int insert = 0;
	int ret;
2963

2964
	if (!root->orphan_block_rsv) {
2965
		block_rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
2966 2967
		if (!block_rsv)
			return -ENOMEM;
2968
	}
2969

2970 2971 2972 2973 2974 2975
	spin_lock(&root->orphan_lock);
	if (!root->orphan_block_rsv) {
		root->orphan_block_rsv = block_rsv;
	} else if (block_rsv) {
		btrfs_free_block_rsv(root, block_rsv);
		block_rsv = NULL;
2976 2977
	}

2978 2979
	if (!test_and_set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			      &BTRFS_I(inode)->runtime_flags)) {
2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991
#if 0
		/*
		 * For proper ENOSPC handling, we should do orphan
		 * cleanup when mounting. But this introduces backward
		 * compatibility issue.
		 */
		if (!xchg(&root->orphan_item_inserted, 1))
			insert = 2;
		else
			insert = 1;
#endif
		insert = 1;
2992
		atomic_inc(&root->orphan_inodes);
2993 2994
	}

2995 2996
	if (!test_and_set_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
			      &BTRFS_I(inode)->runtime_flags))
2997 2998
		reserve = 1;
	spin_unlock(&root->orphan_lock);
2999

3000 3001 3002
	/* grab metadata reservation from transaction handle */
	if (reserve) {
		ret = btrfs_orphan_reserve_metadata(trans, inode);
3003
		BUG_ON(ret); /* -ENOSPC in reservation; Logic error? JDM */
3004
	}
3005

3006 3007
	/* insert an orphan item to track this unlinked/truncated file */
	if (insert >= 1) {
L
Li Zefan 已提交
3008
		ret = btrfs_insert_orphan_item(trans, root, btrfs_ino(inode));
3009
		if (ret && ret != -EEXIST) {
3010 3011
			clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
				  &BTRFS_I(inode)->runtime_flags);
3012 3013 3014 3015
			btrfs_abort_transaction(trans, root, ret);
			return ret;
		}
		ret = 0;
3016 3017 3018 3019 3020 3021
	}

	/* insert an orphan item to track subvolume contains orphan files */
	if (insert >= 2) {
		ret = btrfs_insert_orphan_item(trans, root->fs_info->tree_root,
					       root->root_key.objectid);
3022 3023 3024 3025
		if (ret && ret != -EEXIST) {
			btrfs_abort_transaction(trans, root, ret);
			return ret;
		}
3026 3027
	}
	return 0;
3028 3029 3030 3031 3032 3033
}

/*
 * We have done the truncate/delete so we can go ahead and remove the orphan
 * item for this particular inode.
 */
3034 3035
static int btrfs_orphan_del(struct btrfs_trans_handle *trans,
			    struct inode *inode)
3036 3037
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
3038 3039
	int delete_item = 0;
	int release_rsv = 0;
3040 3041
	int ret = 0;

3042
	spin_lock(&root->orphan_lock);
3043 3044
	if (test_and_clear_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			       &BTRFS_I(inode)->runtime_flags))
3045
		delete_item = 1;
3046

3047 3048
	if (test_and_clear_bit(BTRFS_INODE_ORPHAN_META_RESERVED,
			       &BTRFS_I(inode)->runtime_flags))
3049 3050
		release_rsv = 1;
	spin_unlock(&root->orphan_lock);
3051

3052
	if (trans && delete_item) {
L
Li Zefan 已提交
3053
		ret = btrfs_del_orphan_item(trans, root, btrfs_ino(inode));
3054
		BUG_ON(ret); /* -ENOMEM or corruption (JDM: Recheck) */
3055
	}
3056

3057
	if (release_rsv) {
3058
		btrfs_orphan_release_metadata(inode);
3059 3060
		atomic_dec(&root->orphan_inodes);
	}
3061

3062
	return 0;
3063 3064 3065 3066 3067 3068
}

/*
 * this cleans up any orphans that may be left on the list from the last use
 * of this root.
 */
3069
int btrfs_orphan_cleanup(struct btrfs_root *root)
3070 3071 3072 3073 3074 3075
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_key key, found_key;
	struct btrfs_trans_handle *trans;
	struct inode *inode;
3076
	u64 last_objectid = 0;
3077 3078
	int ret = 0, nr_unlink = 0, nr_truncate = 0;

3079
	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
3080
		return 0;
3081 3082

	path = btrfs_alloc_path();
3083 3084 3085 3086
	if (!path) {
		ret = -ENOMEM;
		goto out;
	}
3087 3088 3089 3090 3091 3092 3093 3094
	path->reada = -1;

	key.objectid = BTRFS_ORPHAN_OBJECTID;
	btrfs_set_key_type(&key, BTRFS_ORPHAN_ITEM_KEY);
	key.offset = (u64)-1;

	while (1) {
		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
3095 3096
		if (ret < 0)
			goto out;
3097 3098 3099

		/*
		 * if ret == 0 means we found what we were searching for, which
L
Lucas De Marchi 已提交
3100
		 * is weird, but possible, so only screw with path if we didn't
3101 3102 3103
		 * find the key and see if we have stuff that matches
		 */
		if (ret > 0) {
3104
			ret = 0;
3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
			if (path->slots[0] == 0)
				break;
			path->slots[0]--;
		}

		/* pull out the item */
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

		/* make sure the item matches what we want */
		if (found_key.objectid != BTRFS_ORPHAN_OBJECTID)
			break;
		if (btrfs_key_type(&found_key) != BTRFS_ORPHAN_ITEM_KEY)
			break;

		/* release the path since we're done with it */
3121
		btrfs_release_path(path);
3122 3123 3124 3125 3126 3127

		/*
		 * this is where we are basically btrfs_lookup, without the
		 * crossing root thing.  we store the inode number in the
		 * offset of the orphan item.
		 */
3128 3129

		if (found_key.offset == last_objectid) {
3130 3131
			btrfs_err(root->fs_info,
				"Error removing orphan entry, stopping orphan cleanup");
3132 3133 3134 3135 3136 3137
			ret = -EINVAL;
			goto out;
		}

		last_objectid = found_key.offset;

3138 3139 3140
		found_key.objectid = found_key.offset;
		found_key.type = BTRFS_INODE_ITEM_KEY;
		found_key.offset = 0;
3141
		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
J
Josef Bacik 已提交
3142 3143
		ret = PTR_RET(inode);
		if (ret && ret != -ESTALE)
3144
			goto out;
3145

3146 3147 3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165 3166 3167 3168 3169 3170 3171 3172 3173 3174 3175 3176 3177
		if (ret == -ESTALE && root == root->fs_info->tree_root) {
			struct btrfs_root *dead_root;
			struct btrfs_fs_info *fs_info = root->fs_info;
			int is_dead_root = 0;

			/*
			 * this is an orphan in the tree root. Currently these
			 * could come from 2 sources:
			 *  a) a snapshot deletion in progress
			 *  b) a free space cache inode
			 * We need to distinguish those two, as the snapshot
			 * orphan must not get deleted.
			 * find_dead_roots already ran before us, so if this
			 * is a snapshot deletion, we should find the root
			 * in the dead_roots list
			 */
			spin_lock(&fs_info->trans_lock);
			list_for_each_entry(dead_root, &fs_info->dead_roots,
					    root_list) {
				if (dead_root->root_key.objectid ==
				    found_key.objectid) {
					is_dead_root = 1;
					break;
				}
			}
			spin_unlock(&fs_info->trans_lock);
			if (is_dead_root) {
				/* prevent this orphan from being found again */
				key.offset = found_key.objectid - 1;
				continue;
			}
		}
3178
		/*
J
Josef Bacik 已提交
3179 3180
		 * Inode is already gone but the orphan item is still there,
		 * kill the orphan item.
3181
		 */
J
Josef Bacik 已提交
3182 3183
		if (ret == -ESTALE) {
			trans = btrfs_start_transaction(root, 1);
3184 3185 3186 3187
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				goto out;
			}
3188 3189
			btrfs_debug(root->fs_info, "auto deleting %Lu",
				found_key.objectid);
J
Josef Bacik 已提交
3190 3191
			ret = btrfs_del_orphan_item(trans, root,
						    found_key.objectid);
3192
			BUG_ON(ret); /* -ENOMEM or corruption (JDM: Recheck) */
3193
			btrfs_end_transaction(trans, root);
3194 3195 3196
			continue;
		}

J
Josef Bacik 已提交
3197 3198 3199 3200
		/*
		 * add this inode to the orphan list so btrfs_orphan_del does
		 * the proper thing when we hit it
		 */
3201 3202
		set_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
			&BTRFS_I(inode)->runtime_flags);
3203
		atomic_inc(&root->orphan_inodes);
J
Josef Bacik 已提交
3204

3205 3206
		/* if we have links, this was a truncate, lets do that */
		if (inode->i_nlink) {
3207 3208 3209 3210 3211
			if (!S_ISREG(inode->i_mode)) {
				WARN_ON(1);
				iput(inode);
				continue;
			}
3212
			nr_truncate++;
3213 3214 3215 3216 3217 3218 3219 3220 3221 3222 3223 3224

			/* 1 for the orphan item deletion. */
			trans = btrfs_start_transaction(root, 1);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				goto out;
			}
			ret = btrfs_orphan_add(trans, inode);
			btrfs_end_transaction(trans, root);
			if (ret)
				goto out;

3225
			ret = btrfs_truncate(inode);
3226 3227
			if (ret)
				btrfs_orphan_del(NULL, inode);
3228 3229 3230 3231 3232 3233
		} else {
			nr_unlink++;
		}

		/* this will do delete_inode and everything for us */
		iput(inode);
3234 3235
		if (ret)
			goto out;
3236
	}
3237 3238 3239
	/* release the path since we're done with it */
	btrfs_release_path(path);

3240 3241 3242 3243 3244 3245 3246
	root->orphan_cleanup_state = ORPHAN_CLEANUP_DONE;

	if (root->orphan_block_rsv)
		btrfs_block_rsv_release(root, root->orphan_block_rsv,
					(u64)-1);

	if (root->orphan_block_rsv || root->orphan_item_inserted) {
3247
		trans = btrfs_join_transaction(root);
3248 3249
		if (!IS_ERR(trans))
			btrfs_end_transaction(trans, root);
3250
	}
3251 3252

	if (nr_unlink)
3253
		btrfs_debug(root->fs_info, "unlinked %d orphans", nr_unlink);
3254
	if (nr_truncate)
3255
		btrfs_debug(root->fs_info, "truncated %d orphans", nr_truncate);
3256 3257 3258

out:
	if (ret)
3259 3260
		btrfs_crit(root->fs_info,
			"could not do orphan cleanup %d", ret);
3261 3262
	btrfs_free_path(path);
	return ret;
3263 3264
}

3265 3266 3267 3268 3269 3270 3271 3272 3273 3274 3275 3276 3277 3278 3279 3280 3281 3282 3283 3284 3285 3286 3287 3288 3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299 3300 3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312 3313 3314 3315
/*
 * very simple check to peek ahead in the leaf looking for xattrs.  If we
 * don't find any xattrs, we know there can't be any acls.
 *
 * slot is the slot the inode is in, objectid is the objectid of the inode
 */
static noinline int acls_after_inode_item(struct extent_buffer *leaf,
					  int slot, u64 objectid)
{
	u32 nritems = btrfs_header_nritems(leaf);
	struct btrfs_key found_key;
	int scanned = 0;

	slot++;
	while (slot < nritems) {
		btrfs_item_key_to_cpu(leaf, &found_key, slot);

		/* we found a different objectid, there must not be acls */
		if (found_key.objectid != objectid)
			return 0;

		/* we found an xattr, assume we've got an acl */
		if (found_key.type == BTRFS_XATTR_ITEM_KEY)
			return 1;

		/*
		 * we found a key greater than an xattr key, there can't
		 * be any acls later on
		 */
		if (found_key.type > BTRFS_XATTR_ITEM_KEY)
			return 0;

		slot++;
		scanned++;

		/*
		 * it goes inode, inode backrefs, xattrs, extents,
		 * so if there are a ton of hard links to an inode there can
		 * be a lot of backrefs.  Don't waste time searching too hard,
		 * this is just an optimization
		 */
		if (scanned >= 8)
			break;
	}
	/* we hit the end of the leaf before we found an xattr or
	 * something larger than an xattr.  We have to assume the inode
	 * has acls
	 */
	return 1;
}

C
Chris Mason 已提交
3316 3317 3318
/*
 * read an inode from the btree into the in-memory inode
 */
3319
static void btrfs_read_locked_inode(struct inode *inode)
C
Chris Mason 已提交
3320 3321
{
	struct btrfs_path *path;
3322
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3323
	struct btrfs_inode_item *inode_item;
3324
	struct btrfs_timespec *tspec;
C
Chris Mason 已提交
3325 3326
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key location;
3327
	int maybe_acls;
J
Josef Bacik 已提交
3328
	u32 rdev;
C
Chris Mason 已提交
3329
	int ret;
3330 3331 3332 3333 3334
	bool filled = false;

	ret = btrfs_fill_inode(inode, &rdev);
	if (!ret)
		filled = true;
C
Chris Mason 已提交
3335 3336

	path = btrfs_alloc_path();
3337 3338 3339
	if (!path)
		goto make_bad;

3340
	path->leave_spinning = 1;
C
Chris Mason 已提交
3341
	memcpy(&location, &BTRFS_I(inode)->location, sizeof(location));
C
Chris Mason 已提交
3342

C
Chris Mason 已提交
3343
	ret = btrfs_lookup_inode(NULL, root, path, &location, 0);
3344
	if (ret)
C
Chris Mason 已提交
3345 3346
		goto make_bad;

3347
	leaf = path->nodes[0];
3348 3349 3350 3351

	if (filled)
		goto cache_acl;

3352 3353 3354
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
				    struct btrfs_inode_item);
	inode->i_mode = btrfs_inode_mode(leaf, inode_item);
M
Miklos Szeredi 已提交
3355
	set_nlink(inode, btrfs_inode_nlink(leaf, inode_item));
3356 3357
	i_uid_write(inode, btrfs_inode_uid(leaf, inode_item));
	i_gid_write(inode, btrfs_inode_gid(leaf, inode_item));
3358
	btrfs_i_size_write(inode, btrfs_inode_size(leaf, inode_item));
3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371

	tspec = btrfs_inode_atime(inode_item);
	inode->i_atime.tv_sec = btrfs_timespec_sec(leaf, tspec);
	inode->i_atime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);

	tspec = btrfs_inode_mtime(inode_item);
	inode->i_mtime.tv_sec = btrfs_timespec_sec(leaf, tspec);
	inode->i_mtime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);

	tspec = btrfs_inode_ctime(inode_item);
	inode->i_ctime.tv_sec = btrfs_timespec_sec(leaf, tspec);
	inode->i_ctime.tv_nsec = btrfs_timespec_nsec(leaf, tspec);

3372
	inode_set_bytes(inode, btrfs_inode_nbytes(leaf, inode_item));
3373
	BTRFS_I(inode)->generation = btrfs_inode_generation(leaf, inode_item);
J
Josef Bacik 已提交
3374 3375 3376 3377 3378 3379 3380 3381 3382 3383 3384 3385
	BTRFS_I(inode)->last_trans = btrfs_inode_transid(leaf, inode_item);

	/*
	 * If we were modified in the current generation and evicted from memory
	 * and then re-read we need to do a full sync since we don't have any
	 * idea about which extents were modified before we were evicted from
	 * cache.
	 */
	if (BTRFS_I(inode)->last_trans == root->fs_info->generation)
		set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
			&BTRFS_I(inode)->runtime_flags);

3386
	inode->i_version = btrfs_inode_sequence(leaf, inode_item);
3387
	inode->i_generation = BTRFS_I(inode)->generation;
J
Josef Bacik 已提交
3388
	inode->i_rdev = 0;
3389 3390
	rdev = btrfs_inode_rdev(leaf, inode_item);

3391
	BTRFS_I(inode)->index_cnt = (u64)-1;
3392
	BTRFS_I(inode)->flags = btrfs_inode_flags(leaf, inode_item);
3393
cache_acl:
3394 3395 3396 3397
	/*
	 * try to precache a NULL acl entry for files that don't have
	 * any xattrs or acls
	 */
L
Li Zefan 已提交
3398 3399
	maybe_acls = acls_after_inode_item(leaf, path->slots[0],
					   btrfs_ino(inode));
3400 3401
	if (!maybe_acls)
		cache_no_acl(inode);
3402

C
Chris Mason 已提交
3403 3404 3405 3406 3407
	btrfs_free_path(path);

	switch (inode->i_mode & S_IFMT) {
	case S_IFREG:
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
3408
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
3409
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
3410 3411 3412 3413 3414 3415 3416 3417 3418 3419 3420 3421 3422
		inode->i_fop = &btrfs_file_operations;
		inode->i_op = &btrfs_file_inode_operations;
		break;
	case S_IFDIR:
		inode->i_fop = &btrfs_dir_file_operations;
		if (root == root->fs_info->tree_root)
			inode->i_op = &btrfs_dir_ro_inode_operations;
		else
			inode->i_op = &btrfs_dir_inode_operations;
		break;
	case S_IFLNK:
		inode->i_op = &btrfs_symlink_inode_operations;
		inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
3423
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
C
Chris Mason 已提交
3424
		break;
J
Josef Bacik 已提交
3425
	default:
J
Jim Owens 已提交
3426
		inode->i_op = &btrfs_special_inode_operations;
J
Josef Bacik 已提交
3427 3428
		init_special_inode(inode, inode->i_mode, rdev);
		break;
C
Chris Mason 已提交
3429
	}
3430 3431

	btrfs_update_iflags(inode);
C
Chris Mason 已提交
3432 3433 3434 3435 3436 3437 3438
	return;

make_bad:
	btrfs_free_path(path);
	make_bad_inode(inode);
}

C
Chris Mason 已提交
3439 3440 3441
/*
 * given a leaf and an inode, copy the inode fields into the leaf
 */
3442 3443
static void fill_inode_item(struct btrfs_trans_handle *trans,
			    struct extent_buffer *leaf,
3444
			    struct btrfs_inode_item *item,
C
Chris Mason 已提交
3445 3446
			    struct inode *inode)
{
3447 3448 3449
	struct btrfs_map_token token;

	btrfs_init_map_token(&token);
3450

3451 3452 3453 3454 3455 3456
	btrfs_set_token_inode_uid(leaf, item, i_uid_read(inode), &token);
	btrfs_set_token_inode_gid(leaf, item, i_gid_read(inode), &token);
	btrfs_set_token_inode_size(leaf, item, BTRFS_I(inode)->disk_i_size,
				   &token);
	btrfs_set_token_inode_mode(leaf, item, inode->i_mode, &token);
	btrfs_set_token_inode_nlink(leaf, item, inode->i_nlink, &token);
3457

3458 3459 3460 3461
	btrfs_set_token_timespec_sec(leaf, btrfs_inode_atime(item),
				     inode->i_atime.tv_sec, &token);
	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_atime(item),
				      inode->i_atime.tv_nsec, &token);
3462

3463 3464 3465 3466
	btrfs_set_token_timespec_sec(leaf, btrfs_inode_mtime(item),
				     inode->i_mtime.tv_sec, &token);
	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_mtime(item),
				      inode->i_mtime.tv_nsec, &token);
3467

3468 3469 3470 3471
	btrfs_set_token_timespec_sec(leaf, btrfs_inode_ctime(item),
				     inode->i_ctime.tv_sec, &token);
	btrfs_set_token_timespec_nsec(leaf, btrfs_inode_ctime(item),
				      inode->i_ctime.tv_nsec, &token);
3472

3473 3474 3475 3476 3477 3478 3479 3480 3481
	btrfs_set_token_inode_nbytes(leaf, item, inode_get_bytes(inode),
				     &token);
	btrfs_set_token_inode_generation(leaf, item, BTRFS_I(inode)->generation,
					 &token);
	btrfs_set_token_inode_sequence(leaf, item, inode->i_version, &token);
	btrfs_set_token_inode_transid(leaf, item, trans->transid, &token);
	btrfs_set_token_inode_rdev(leaf, item, inode->i_rdev, &token);
	btrfs_set_token_inode_flags(leaf, item, BTRFS_I(inode)->flags, &token);
	btrfs_set_token_inode_block_group(leaf, item, 0, &token);
C
Chris Mason 已提交
3482 3483
}

C
Chris Mason 已提交
3484 3485 3486
/*
 * copy everything in the in-memory inode into the btree.
 */
3487
static noinline int btrfs_update_inode_item(struct btrfs_trans_handle *trans,
C
Chris Mason 已提交
3488
				struct btrfs_root *root, struct inode *inode)
C
Chris Mason 已提交
3489 3490 3491
{
	struct btrfs_inode_item *inode_item;
	struct btrfs_path *path;
3492
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3493 3494 3495
	int ret;

	path = btrfs_alloc_path();
3496 3497 3498
	if (!path)
		return -ENOMEM;

3499
	path->leave_spinning = 1;
3500 3501
	ret = btrfs_lookup_inode(trans, root, path, &BTRFS_I(inode)->location,
				 1);
C
Chris Mason 已提交
3502 3503 3504 3505 3506 3507
	if (ret) {
		if (ret > 0)
			ret = -ENOENT;
		goto failed;
	}

3508
	btrfs_unlock_up_safe(path, 1);
3509 3510
	leaf = path->nodes[0];
	inode_item = btrfs_item_ptr(leaf, path->slots[0],
3511
				    struct btrfs_inode_item);
C
Chris Mason 已提交
3512

3513
	fill_inode_item(trans, leaf, inode_item, inode);
3514
	btrfs_mark_buffer_dirty(leaf);
3515
	btrfs_set_inode_last_trans(trans, inode);
C
Chris Mason 已提交
3516 3517 3518 3519 3520 3521
	ret = 0;
failed:
	btrfs_free_path(path);
	return ret;
}

3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536
/*
 * copy everything in the in-memory inode into the btree.
 */
noinline int btrfs_update_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root, struct inode *inode)
{
	int ret;

	/*
	 * If the inode is a free space inode, we can deadlock during commit
	 * if we put it into the delayed code.
	 *
	 * The data relocation inode should also be directly updated
	 * without delay
	 */
3537
	if (!btrfs_is_free_space_inode(inode)
3538
	    && root->root_key.objectid != BTRFS_DATA_RELOC_TREE_OBJECTID) {
3539 3540
		btrfs_update_root_times(trans, root);

3541 3542 3543 3544 3545 3546 3547 3548 3549
		ret = btrfs_delayed_update_inode(trans, root, inode);
		if (!ret)
			btrfs_set_inode_last_trans(trans, inode);
		return ret;
	}

	return btrfs_update_inode_item(trans, root, inode);
}

3550 3551 3552
noinline int btrfs_update_inode_fallback(struct btrfs_trans_handle *trans,
					 struct btrfs_root *root,
					 struct inode *inode)
3553 3554 3555 3556 3557 3558 3559 3560 3561
{
	int ret;

	ret = btrfs_update_inode(trans, root, inode);
	if (ret == -ENOSPC)
		return btrfs_update_inode_item(trans, root, inode);
	return ret;
}

C
Chris Mason 已提交
3562 3563 3564 3565 3566
/*
 * unlink helper that gets used here in inode.c and in the tree logging
 * recovery code.  It remove a link in a directory with a given name, and
 * also drops the back refs in the inode to the directory
 */
3567 3568 3569 3570
static int __btrfs_unlink_inode(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct inode *dir, struct inode *inode,
				const char *name, int name_len)
C
Chris Mason 已提交
3571 3572 3573
{
	struct btrfs_path *path;
	int ret = 0;
3574
	struct extent_buffer *leaf;
C
Chris Mason 已提交
3575
	struct btrfs_dir_item *di;
3576
	struct btrfs_key key;
3577
	u64 index;
L
Li Zefan 已提交
3578 3579
	u64 ino = btrfs_ino(inode);
	u64 dir_ino = btrfs_ino(dir);
C
Chris Mason 已提交
3580 3581

	path = btrfs_alloc_path();
3582 3583
	if (!path) {
		ret = -ENOMEM;
3584
		goto out;
3585 3586
	}

3587
	path->leave_spinning = 1;
L
Li Zefan 已提交
3588
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
C
Chris Mason 已提交
3589 3590 3591 3592 3593 3594 3595 3596 3597
				    name, name_len, -1);
	if (IS_ERR(di)) {
		ret = PTR_ERR(di);
		goto err;
	}
	if (!di) {
		ret = -ENOENT;
		goto err;
	}
3598 3599
	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
C
Chris Mason 已提交
3600
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
3601 3602
	if (ret)
		goto err;
3603
	btrfs_release_path(path);
C
Chris Mason 已提交
3604

L
Li Zefan 已提交
3605 3606
	ret = btrfs_del_inode_ref(trans, root, name, name_len, ino,
				  dir_ino, &index);
3607
	if (ret) {
3608 3609 3610 3611
		btrfs_info(root->fs_info,
			"failed to delete reference to %.*s, inode %llu parent %llu",
			name_len, name,
			(unsigned long long)ino, (unsigned long long)dir_ino);
3612
		btrfs_abort_transaction(trans, root, ret);
3613 3614 3615
		goto err;
	}

3616
	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
3617 3618
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
3619
		goto err;
3620
	}
C
Chris Mason 已提交
3621

3622
	ret = btrfs_del_inode_ref_in_log(trans, root, name, name_len,
L
Li Zefan 已提交
3623
					 inode, dir_ino);
3624 3625 3626 3627
	if (ret != 0 && ret != -ENOENT) {
		btrfs_abort_transaction(trans, root, ret);
		goto err;
	}
3628 3629 3630

	ret = btrfs_del_dir_entries_in_log(trans, root, name, name_len,
					   dir, index);
3631 3632
	if (ret == -ENOENT)
		ret = 0;
3633 3634
	else if (ret)
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
3635 3636
err:
	btrfs_free_path(path);
3637 3638 3639 3640
	if (ret)
		goto out;

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
3641 3642
	inode_inc_iversion(inode);
	inode_inc_iversion(dir);
3643
	inode->i_ctime = dir->i_mtime = dir->i_ctime = CURRENT_TIME;
3644
	ret = btrfs_update_inode(trans, root, dir);
3645
out:
C
Chris Mason 已提交
3646 3647 3648
	return ret;
}

3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661 3662 3663
int btrfs_unlink_inode(struct btrfs_trans_handle *trans,
		       struct btrfs_root *root,
		       struct inode *dir, struct inode *inode,
		       const char *name, int name_len)
{
	int ret;
	ret = __btrfs_unlink_inode(trans, root, dir, inode, name, name_len);
	if (!ret) {
		btrfs_drop_nlink(inode);
		ret = btrfs_update_inode(trans, root, inode);
	}
	return ret;
}
		

3664 3665 3666
/* helper to check if there is any shared block in the path */
static int check_path_shared(struct btrfs_root *root,
			     struct btrfs_path *path)
C
Chris Mason 已提交
3667
{
3668 3669
	struct extent_buffer *eb;
	int level;
3670
	u64 refs = 1;
3671

3672
	for (level = 0; level < BTRFS_MAX_LEVEL; level++) {
3673 3674
		int ret;

3675 3676 3677 3678 3679
		if (!path->nodes[level])
			break;
		eb = path->nodes[level];
		if (!btrfs_block_can_be_shared(root, eb))
			continue;
3680
		ret = btrfs_lookup_extent_info(NULL, root, eb->start, level, 1,
3681 3682 3683
					       &refs, NULL);
		if (refs > 1)
			return 1;
3684
	}
3685
	return 0;
C
Chris Mason 已提交
3686 3687
}

3688 3689 3690 3691 3692 3693 3694 3695 3696
/*
 * helper to start transaction for unlink and rmdir.
 *
 * unlink and rmdir are special in btrfs, they do not always free space.
 * so in enospc case, we should make sure they will free space before
 * allowing them to use the global metadata reservation.
 */
static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
						       struct dentry *dentry)
3697
{
C
Chris Mason 已提交
3698
	struct btrfs_trans_handle *trans;
3699
	struct btrfs_root *root = BTRFS_I(dir)->root;
3700 3701
	struct btrfs_path *path;
	struct btrfs_dir_item *di;
3702
	struct inode *inode = dentry->d_inode;
3703
	u64 index;
3704 3705
	int check_link = 1;
	int err = -ENOSPC;
3706
	int ret;
L
Li Zefan 已提交
3707 3708
	u64 ino = btrfs_ino(inode);
	u64 dir_ino = btrfs_ino(dir);
3709

3710 3711 3712 3713 3714 3715 3716
	/*
	 * 1 for the possible orphan item
	 * 1 for the dir item
	 * 1 for the dir index
	 * 1 for the inode ref
	 * 1 for the inode
	 */
3717
	trans = btrfs_start_transaction(root, 5);
3718 3719
	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
		return trans;
3720

L
Li Zefan 已提交
3721
	if (ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
3722
		return ERR_PTR(-ENOSPC);
3723

3724 3725 3726
	/* check if there is someone else holds reference */
	if (S_ISDIR(inode->i_mode) && atomic_read(&inode->i_count) > 1)
		return ERR_PTR(-ENOSPC);
3727

3728 3729
	if (atomic_read(&inode->i_count) > 2)
		return ERR_PTR(-ENOSPC);
3730

3731 3732 3733 3734 3735 3736 3737
	if (xchg(&root->fs_info->enospc_unlink, 1))
		return ERR_PTR(-ENOSPC);

	path = btrfs_alloc_path();
	if (!path) {
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(-ENOMEM);
3738 3739
	}

3740 3741
	/* 1 for the orphan item */
	trans = btrfs_start_transaction(root, 1);
3742
	if (IS_ERR(trans)) {
3743 3744 3745 3746
		btrfs_free_path(path);
		root->fs_info->enospc_unlink = 0;
		return trans;
	}
3747

3748 3749
	path->skip_locking = 1;
	path->search_commit_root = 1;
3750

3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761
	ret = btrfs_lookup_inode(trans, root, path,
				&BTRFS_I(dir)->location, 0);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret == 0) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		check_link = 0;
3762
	}
3763
	btrfs_release_path(path);
3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776

	ret = btrfs_lookup_inode(trans, root, path,
				&BTRFS_I(inode)->location, 0);
	if (ret < 0) {
		err = ret;
		goto out;
	}
	if (ret == 0) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		check_link = 0;
	}
3777
	btrfs_release_path(path);
3778 3779 3780

	if (ret == 0 && S_ISREG(inode->i_mode)) {
		ret = btrfs_lookup_file_extent(trans, root, path,
L
Li Zefan 已提交
3781
					       ino, (u64)-1, 0);
3782 3783 3784 3785
		if (ret < 0) {
			err = ret;
			goto out;
		}
3786
		BUG_ON(ret == 0); /* Corruption */
3787 3788
		if (check_path_shared(root, path))
			goto out;
3789
		btrfs_release_path(path);
3790 3791 3792 3793 3794 3795 3796
	}

	if (!check_link) {
		err = 0;
		goto out;
	}

L
Li Zefan 已提交
3797
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
3798 3799 3800 3801 3802 3803 3804 3805 3806 3807 3808 3809
				dentry->d_name.name, dentry->d_name.len, 0);
	if (IS_ERR(di)) {
		err = PTR_ERR(di);
		goto out;
	}
	if (di) {
		if (check_path_shared(root, path))
			goto out;
	} else {
		err = 0;
		goto out;
	}
3810
	btrfs_release_path(path);
3811

M
Mark Fasheh 已提交
3812 3813 3814 3815 3816
	ret = btrfs_get_inode_ref_index(trans, root, path, dentry->d_name.name,
					dentry->d_name.len, ino, dir_ino, 0,
					&index);
	if (ret) {
		err = ret;
3817 3818
		goto out;
	}
M
Mark Fasheh 已提交
3819

3820 3821
	if (check_path_shared(root, path))
		goto out;
M
Mark Fasheh 已提交
3822

3823
	btrfs_release_path(path);
3824

3825 3826 3827 3828 3829 3830 3831 3832
	/*
	 * This is a commit root search, if we can lookup inode item and other
	 * relative items in the commit root, it means the transaction of
	 * dir/file creation has been committed, and the dir index item that we
	 * delay to insert has also been inserted into the commit root. So
	 * we needn't worry about the delayed insertion of the dir index item
	 * here.
	 */
L
Li Zefan 已提交
3833
	di = btrfs_lookup_dir_index_item(trans, root, path, dir_ino, index,
3834 3835 3836 3837 3838 3839 3840 3841 3842 3843 3844 3845
				dentry->d_name.name, dentry->d_name.len, 0);
	if (IS_ERR(di)) {
		err = PTR_ERR(di);
		goto out;
	}
	BUG_ON(ret == -ENOENT);
	if (check_path_shared(root, path))
		goto out;

	err = 0;
out:
	btrfs_free_path(path);
3846 3847 3848 3849
	/* Migrate the orphan reservation over */
	if (!err)
		err = btrfs_block_rsv_migrate(trans->block_rsv,
				&root->fs_info->global_block_rsv,
3850
				trans->bytes_reserved);
3851

3852 3853 3854 3855 3856 3857 3858 3859 3860 3861 3862 3863 3864
	if (err) {
		btrfs_end_transaction(trans, root);
		root->fs_info->enospc_unlink = 0;
		return ERR_PTR(err);
	}

	trans->block_rsv = &root->fs_info->global_block_rsv;
	return trans;
}

static void __unlink_end_trans(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root)
{
3865
	if (trans->block_rsv->type == BTRFS_BLOCK_RSV_GLOBAL) {
3866 3867 3868
		btrfs_block_rsv_release(root, trans->block_rsv,
					trans->bytes_reserved);
		trans->block_rsv = &root->fs_info->trans_block_rsv;
3869 3870 3871
		BUG_ON(!root->fs_info->enospc_unlink);
		root->fs_info->enospc_unlink = 0;
	}
3872
	btrfs_end_transaction(trans, root);
3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884
}

static int btrfs_unlink(struct inode *dir, struct dentry *dentry)
{
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;
	struct inode *inode = dentry->d_inode;
	int ret;

	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
3885

3886 3887
	btrfs_record_unlink_dir(trans, dir, dentry->d_inode, 0);

3888 3889
	ret = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
3890 3891
	if (ret)
		goto out;
3892

3893
	if (inode->i_nlink == 0) {
3894
		ret = btrfs_orphan_add(trans, inode);
3895 3896
		if (ret)
			goto out;
3897
	}
3898

3899
out:
3900
	__unlink_end_trans(trans, root);
3901
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
3902 3903 3904
	return ret;
}

3905 3906 3907 3908 3909 3910 3911 3912 3913 3914 3915
int btrfs_unlink_subvol(struct btrfs_trans_handle *trans,
			struct btrfs_root *root,
			struct inode *dir, u64 objectid,
			const char *name, int name_len)
{
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
	u64 index;
	int ret;
L
Li Zefan 已提交
3916
	u64 dir_ino = btrfs_ino(dir);
3917 3918 3919 3920 3921

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

L
Li Zefan 已提交
3922
	di = btrfs_lookup_dir_item(trans, root, path, dir_ino,
3923
				   name, name_len, -1);
3924 3925 3926 3927 3928 3929 3930
	if (IS_ERR_OR_NULL(di)) {
		if (!di)
			ret = -ENOENT;
		else
			ret = PTR_ERR(di);
		goto out;
	}
3931 3932 3933 3934 3935

	leaf = path->nodes[0];
	btrfs_dir_item_key_to_cpu(leaf, di, &key);
	WARN_ON(key.type != BTRFS_ROOT_ITEM_KEY || key.objectid != objectid);
	ret = btrfs_delete_one_dir_name(trans, root, path, di);
3936 3937 3938 3939
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out;
	}
3940
	btrfs_release_path(path);
3941 3942 3943

	ret = btrfs_del_root_ref(trans, root->fs_info->tree_root,
				 objectid, root->root_key.objectid,
L
Li Zefan 已提交
3944
				 dir_ino, &index, name, name_len);
3945
	if (ret < 0) {
3946 3947 3948 3949
		if (ret != -ENOENT) {
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
L
Li Zefan 已提交
3950
		di = btrfs_search_dir_index_item(root, path, dir_ino,
3951
						 name, name_len);
3952 3953 3954 3955 3956 3957 3958 3959
		if (IS_ERR_OR_NULL(di)) {
			if (!di)
				ret = -ENOENT;
			else
				ret = PTR_ERR(di);
			btrfs_abort_transaction(trans, root, ret);
			goto out;
		}
3960 3961 3962

		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
3963
		btrfs_release_path(path);
3964 3965
		index = key.offset;
	}
3966
	btrfs_release_path(path);
3967

3968
	ret = btrfs_delete_delayed_dir_index(trans, root, dir, index);
3969 3970 3971 3972
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out;
	}
3973 3974

	btrfs_i_size_write(dir, dir->i_size - name_len * 2);
3975
	inode_inc_iversion(dir);
3976
	dir->i_mtime = dir->i_ctime = CURRENT_TIME;
3977
	ret = btrfs_update_inode_fallback(trans, root, dir);
3978 3979 3980
	if (ret)
		btrfs_abort_transaction(trans, root, ret);
out:
3981
	btrfs_free_path(path);
3982
	return ret;
3983 3984
}

C
Chris Mason 已提交
3985 3986 3987
static int btrfs_rmdir(struct inode *dir, struct dentry *dentry)
{
	struct inode *inode = dentry->d_inode;
3988
	int err = 0;
C
Chris Mason 已提交
3989 3990 3991
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_trans_handle *trans;

3992
	if (inode->i_size > BTRFS_EMPTY_DIR_SIZE)
Y
Yan 已提交
3993
		return -ENOTEMPTY;
3994 3995
	if (btrfs_ino(inode) == BTRFS_FIRST_FREE_OBJECTID)
		return -EPERM;
Y
Yan 已提交
3996

3997 3998
	trans = __unlink_start_trans(dir, dentry);
	if (IS_ERR(trans))
3999 4000
		return PTR_ERR(trans);

L
Li Zefan 已提交
4001
	if (unlikely(btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
4002 4003 4004 4005 4006 4007 4008
		err = btrfs_unlink_subvol(trans, root, dir,
					  BTRFS_I(inode)->location.objectid,
					  dentry->d_name.name,
					  dentry->d_name.len);
		goto out;
	}

4009 4010
	err = btrfs_orphan_add(trans, inode);
	if (err)
4011
		goto out;
4012

C
Chris Mason 已提交
4013
	/* now the directory is empty */
4014 4015
	err = btrfs_unlink_inode(trans, root, dir, dentry->d_inode,
				 dentry->d_name.name, dentry->d_name.len);
C
Chris Mason 已提交
4016
	if (!err)
4017
		btrfs_i_size_write(inode, 0);
4018
out:
4019
	__unlink_end_trans(trans, root);
4020
	btrfs_btree_balance_dirty(root);
4021

C
Chris Mason 已提交
4022 4023 4024 4025 4026 4027
	return err;
}

/*
 * this can truncate away extent items, csum items and directory items.
 * It starts at a high offset and removes keys until it can't find
C
Chris Mason 已提交
4028
 * any higher than new_size
C
Chris Mason 已提交
4029 4030 4031
 *
 * csum items that cross the new i_size are truncated to the new size
 * as well.
4032 4033 4034
 *
 * min_type is the minimum key type to truncate down to.  If set to 0, this
 * will kill all the items on this inode, including the INODE_ITEM_KEY.
C
Chris Mason 已提交
4035
 */
4036 4037 4038 4039
int btrfs_truncate_inode_items(struct btrfs_trans_handle *trans,
			       struct btrfs_root *root,
			       struct inode *inode,
			       u64 new_size, u32 min_type)
C
Chris Mason 已提交
4040 4041
{
	struct btrfs_path *path;
4042
	struct extent_buffer *leaf;
C
Chris Mason 已提交
4043
	struct btrfs_file_extent_item *fi;
4044 4045
	struct btrfs_key key;
	struct btrfs_key found_key;
C
Chris Mason 已提交
4046
	u64 extent_start = 0;
4047
	u64 extent_num_bytes = 0;
4048
	u64 extent_offset = 0;
C
Chris Mason 已提交
4049
	u64 item_end = 0;
4050
	u32 found_type = (u8)-1;
C
Chris Mason 已提交
4051 4052
	int found_extent;
	int del_item;
4053 4054
	int pending_del_nr = 0;
	int pending_del_slot = 0;
4055
	int extent_type = -1;
4056 4057
	int ret;
	int err = 0;
L
Li Zefan 已提交
4058
	u64 ino = btrfs_ino(inode);
4059 4060

	BUG_ON(new_size > 0 && min_type != BTRFS_EXTENT_DATA_KEY);
C
Chris Mason 已提交
4061

4062 4063 4064 4065 4066
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;
	path->reada = -1;

J
Josef Bacik 已提交
4067 4068 4069 4070 4071
	/*
	 * We want to drop from the next block forward in case this new size is
	 * not block aligned since we will be keeping the last block of the
	 * extent just the way it is.
	 */
4072
	if (root->ref_cows || root == root->fs_info->tree_root)
4073 4074
		btrfs_drop_extent_cache(inode, ALIGN(new_size,
					root->sectorsize), (u64)-1, 0);
4075

4076 4077 4078 4079 4080 4081 4082 4083 4084
	/*
	 * This function is also used to drop the items in the log tree before
	 * we relog the inode, so if root != BTRFS_I(inode)->root, it means
	 * it is used to drop the loged items. So we shouldn't kill the delayed
	 * items.
	 */
	if (min_type == 0 && root == BTRFS_I(inode)->root)
		btrfs_kill_delayed_inode_items(inode);

L
Li Zefan 已提交
4085
	key.objectid = ino;
C
Chris Mason 已提交
4086
	key.offset = (u64)-1;
4087 4088
	key.type = (u8)-1;

4089
search_again:
4090
	path->leave_spinning = 1;
4091
	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
4092 4093 4094 4095
	if (ret < 0) {
		err = ret;
		goto out;
	}
C
Chris Mason 已提交
4096

4097
	if (ret > 0) {
4098 4099 4100
		/* there are no items in the tree for us to truncate, we're
		 * done
		 */
4101 4102
		if (path->slots[0] == 0)
			goto out;
4103 4104 4105
		path->slots[0]--;
	}

C
Chris Mason 已提交
4106
	while (1) {
C
Chris Mason 已提交
4107
		fi = NULL;
4108 4109 4110
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		found_type = btrfs_key_type(&found_key);
C
Chris Mason 已提交
4111

L
Li Zefan 已提交
4112
		if (found_key.objectid != ino)
C
Chris Mason 已提交
4113
			break;
4114

4115
		if (found_type < min_type)
C
Chris Mason 已提交
4116 4117
			break;

4118
		item_end = found_key.offset;
C
Chris Mason 已提交
4119
		if (found_type == BTRFS_EXTENT_DATA_KEY) {
4120
			fi = btrfs_item_ptr(leaf, path->slots[0],
C
Chris Mason 已提交
4121
					    struct btrfs_file_extent_item);
4122 4123
			extent_type = btrfs_file_extent_type(leaf, fi);
			if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
4124
				item_end +=
4125
				    btrfs_file_extent_num_bytes(leaf, fi);
4126 4127
			} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
				item_end += btrfs_file_extent_inline_len(leaf,
C
Chris Mason 已提交
4128
									 fi);
C
Chris Mason 已提交
4129
			}
4130
			item_end--;
C
Chris Mason 已提交
4131
		}
4132 4133 4134 4135
		if (found_type > min_type) {
			del_item = 1;
		} else {
			if (item_end < new_size)
4136
				break;
4137 4138 4139 4140
			if (found_key.offset >= new_size)
				del_item = 1;
			else
				del_item = 0;
C
Chris Mason 已提交
4141 4142 4143
		}
		found_extent = 0;
		/* FIXME, shrink the extent if the ref count is only 1 */
4144 4145 4146 4147
		if (found_type != BTRFS_EXTENT_DATA_KEY)
			goto delete;

		if (extent_type != BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
4148
			u64 num_dec;
4149
			extent_start = btrfs_file_extent_disk_bytenr(leaf, fi);
4150
			if (!del_item) {
4151 4152
				u64 orig_num_bytes =
					btrfs_file_extent_num_bytes(leaf, fi);
4153 4154 4155
				extent_num_bytes = ALIGN(new_size -
						found_key.offset,
						root->sectorsize);
4156 4157 4158
				btrfs_set_file_extent_num_bytes(leaf, fi,
							 extent_num_bytes);
				num_dec = (orig_num_bytes -
C
Chris Mason 已提交
4159
					   extent_num_bytes);
4160
				if (root->ref_cows && extent_start != 0)
4161
					inode_sub_bytes(inode, num_dec);
4162
				btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
4163
			} else {
4164 4165 4166
				extent_num_bytes =
					btrfs_file_extent_disk_num_bytes(leaf,
									 fi);
4167 4168 4169
				extent_offset = found_key.offset -
					btrfs_file_extent_offset(leaf, fi);

C
Chris Mason 已提交
4170
				/* FIXME blocksize != 4096 */
C
Chris Mason 已提交
4171
				num_dec = btrfs_file_extent_num_bytes(leaf, fi);
C
Chris Mason 已提交
4172 4173
				if (extent_start != 0) {
					found_extent = 1;
4174
					if (root->ref_cows)
4175
						inode_sub_bytes(inode, num_dec);
4176
				}
C
Chris Mason 已提交
4177
			}
C
Chris Mason 已提交
4178
		} else if (extent_type == BTRFS_FILE_EXTENT_INLINE) {
C
Chris Mason 已提交
4179 4180 4181 4182 4183 4184 4185 4186
			/*
			 * we can't truncate inline items that have had
			 * special encodings
			 */
			if (!del_item &&
			    btrfs_file_extent_compression(leaf, fi) == 0 &&
			    btrfs_file_extent_encryption(leaf, fi) == 0 &&
			    btrfs_file_extent_other_encoding(leaf, fi) == 0) {
4187 4188 4189
				u32 size = new_size - found_key.offset;

				if (root->ref_cows) {
4190 4191
					inode_sub_bytes(inode, item_end + 1 -
							new_size);
4192 4193 4194
				}
				size =
				    btrfs_file_extent_calc_inline_size(size);
4195
				btrfs_truncate_item(root, path, size, 1);
4196
			} else if (root->ref_cows) {
4197 4198
				inode_sub_bytes(inode, item_end + 1 -
						found_key.offset);
C
Chris Mason 已提交
4199
			}
C
Chris Mason 已提交
4200
		}
4201
delete:
C
Chris Mason 已提交
4202
		if (del_item) {
4203 4204 4205 4206 4207 4208 4209 4210 4211 4212
			if (!pending_del_nr) {
				/* no pending yet, add ourselves */
				pending_del_slot = path->slots[0];
				pending_del_nr = 1;
			} else if (pending_del_nr &&
				   path->slots[0] + 1 == pending_del_slot) {
				/* hop on the pending chunk */
				pending_del_nr++;
				pending_del_slot = path->slots[0];
			} else {
C
Chris Mason 已提交
4213
				BUG();
4214
			}
C
Chris Mason 已提交
4215 4216 4217
		} else {
			break;
		}
4218 4219
		if (found_extent && (root->ref_cows ||
				     root == root->fs_info->tree_root)) {
4220
			btrfs_set_path_blocking(path);
C
Chris Mason 已提交
4221
			ret = btrfs_free_extent(trans, root, extent_start,
4222 4223
						extent_num_bytes, 0,
						btrfs_header_owner(leaf),
A
Arne Jansen 已提交
4224
						ino, extent_offset, 0);
C
Chris Mason 已提交
4225 4226
			BUG_ON(ret);
		}
4227

4228 4229 4230 4231 4232 4233 4234 4235 4236
		if (found_type == BTRFS_INODE_ITEM_KEY)
			break;

		if (path->slots[0] == 0 ||
		    path->slots[0] != pending_del_slot) {
			if (pending_del_nr) {
				ret = btrfs_del_items(trans, root, path,
						pending_del_slot,
						pending_del_nr);
4237 4238 4239 4240 4241
				if (ret) {
					btrfs_abort_transaction(trans,
								root, ret);
					goto error;
				}
4242 4243
				pending_del_nr = 0;
			}
4244
			btrfs_release_path(path);
4245
			goto search_again;
4246 4247
		} else {
			path->slots[0]--;
4248
		}
C
Chris Mason 已提交
4249
	}
4250
out:
4251 4252 4253
	if (pending_del_nr) {
		ret = btrfs_del_items(trans, root, path, pending_del_slot,
				      pending_del_nr);
4254 4255
		if (ret)
			btrfs_abort_transaction(trans, root, ret);
4256
	}
4257
error:
C
Chris Mason 已提交
4258
	btrfs_free_path(path);
4259
	return err;
C
Chris Mason 已提交
4260 4261 4262
}

/*
J
Josef Bacik 已提交
4263 4264 4265 4266 4267 4268 4269 4270 4271
 * btrfs_truncate_page - read, zero a chunk and write a page
 * @inode - inode that we're zeroing
 * @from - the offset to start zeroing
 * @len - the length to zero, 0 to zero the entire range respective to the
 *	offset
 * @front - zero up to the offset instead of from the offset on
 *
 * This will find the page for the "from" offset and cow the page and zero the
 * part we want to zero.  This is used with truncate and hole punching.
C
Chris Mason 已提交
4272
 */
J
Josef Bacik 已提交
4273 4274
int btrfs_truncate_page(struct inode *inode, loff_t from, loff_t len,
			int front)
C
Chris Mason 已提交
4275
{
J
Josef Bacik 已提交
4276
	struct address_space *mapping = inode->i_mapping;
4277
	struct btrfs_root *root = BTRFS_I(inode)->root;
4278 4279
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
4280
	struct extent_state *cached_state = NULL;
4281
	char *kaddr;
4282
	u32 blocksize = root->sectorsize;
C
Chris Mason 已提交
4283 4284 4285
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	struct page *page;
4286
	gfp_t mask = btrfs_alloc_write_mask(mapping);
C
Chris Mason 已提交
4287
	int ret = 0;
4288
	u64 page_start;
4289
	u64 page_end;
C
Chris Mason 已提交
4290

J
Josef Bacik 已提交
4291 4292
	if ((offset & (blocksize - 1)) == 0 &&
	    (!len || ((len & (blocksize - 1)) == 0)))
C
Chris Mason 已提交
4293
		goto out;
4294
	ret = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
4295 4296
	if (ret)
		goto out;
C
Chris Mason 已提交
4297

4298
again:
4299
	page = find_or_create_page(mapping, index, mask);
4300
	if (!page) {
4301
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
4302
		ret = -ENOMEM;
C
Chris Mason 已提交
4303
		goto out;
4304
	}
4305 4306 4307 4308

	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;

C
Chris Mason 已提交
4309
	if (!PageUptodate(page)) {
C
Chris Mason 已提交
4310
		ret = btrfs_readpage(NULL, page);
C
Chris Mason 已提交
4311
		lock_page(page);
4312 4313 4314 4315 4316
		if (page->mapping != mapping) {
			unlock_page(page);
			page_cache_release(page);
			goto again;
		}
C
Chris Mason 已提交
4317 4318
		if (!PageUptodate(page)) {
			ret = -EIO;
4319
			goto out_unlock;
C
Chris Mason 已提交
4320 4321
		}
	}
4322
	wait_on_page_writeback(page);
4323

4324
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
4325 4326 4327 4328
	set_page_extent_mapped(page);

	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
4329 4330
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
4331 4332
		unlock_page(page);
		page_cache_release(page);
4333
		btrfs_start_ordered_extent(inode, ordered, 1);
4334 4335 4336 4337
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

4338
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
4339 4340
			  EXTENT_DIRTY | EXTENT_DELALLOC |
			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
4341
			  0, 0, &cached_state, GFP_NOFS);
4342

4343 4344
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
4345
	if (ret) {
4346 4347
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
4348 4349 4350
		goto out_unlock;
	}

4351
	if (offset != PAGE_CACHE_SIZE) {
J
Josef Bacik 已提交
4352 4353
		if (!len)
			len = PAGE_CACHE_SIZE - offset;
4354
		kaddr = kmap(page);
J
Josef Bacik 已提交
4355 4356 4357 4358
		if (front)
			memset(kaddr, 0, offset);
		else
			memset(kaddr + offset, 0, len);
4359 4360 4361
		flush_dcache_page(page);
		kunmap(page);
	}
4362
	ClearPageChecked(page);
4363
	set_page_dirty(page);
4364 4365
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state,
			     GFP_NOFS);
C
Chris Mason 已提交
4366

4367
out_unlock:
4368
	if (ret)
4369
		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
C
Chris Mason 已提交
4370 4371 4372 4373 4374 4375
	unlock_page(page);
	page_cache_release(page);
out:
	return ret;
}

4376 4377 4378 4379 4380 4381
/*
 * This function puts in dummy file extents for the area we're creating a hole
 * for.  So if we are truncating this file to a larger size we need to insert
 * these file extents so that btrfs_get_extent will return a EXTENT_MAP_HOLE for
 * the range between oldsize and size
 */
4382
int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size)
C
Chris Mason 已提交
4383
{
Y
Yan Zheng 已提交
4384 4385 4386
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
4387
	struct extent_map *em = NULL;
4388
	struct extent_state *cached_state = NULL;
J
Josef Bacik 已提交
4389
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
4390 4391
	u64 hole_start = ALIGN(oldsize, root->sectorsize);
	u64 block_end = ALIGN(size, root->sectorsize);
Y
Yan Zheng 已提交
4392 4393 4394
	u64 last_byte;
	u64 cur_offset;
	u64 hole_size;
J
Josef Bacik 已提交
4395
	int err = 0;
C
Chris Mason 已提交
4396

Y
Yan Zheng 已提交
4397 4398 4399 4400 4401 4402 4403
	if (size <= hole_start)
		return 0;

	while (1) {
		struct btrfs_ordered_extent *ordered;
		btrfs_wait_ordered_range(inode, hole_start,
					 block_end - hole_start);
4404
		lock_extent_bits(io_tree, hole_start, block_end - 1, 0,
4405
				 &cached_state);
Y
Yan Zheng 已提交
4406 4407 4408
		ordered = btrfs_lookup_ordered_extent(inode, hole_start);
		if (!ordered)
			break;
4409 4410
		unlock_extent_cached(io_tree, hole_start, block_end - 1,
				     &cached_state, GFP_NOFS);
Y
Yan Zheng 已提交
4411 4412
		btrfs_put_ordered_extent(ordered);
	}
C
Chris Mason 已提交
4413

Y
Yan Zheng 已提交
4414 4415 4416 4417
	cur_offset = hole_start;
	while (1) {
		em = btrfs_get_extent(inode, NULL, 0, cur_offset,
				block_end - cur_offset, 0);
4418 4419
		if (IS_ERR(em)) {
			err = PTR_ERR(em);
4420
			em = NULL;
4421 4422
			break;
		}
Y
Yan Zheng 已提交
4423
		last_byte = min(extent_map_end(em), block_end);
4424
		last_byte = ALIGN(last_byte , root->sectorsize);
4425
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags)) {
J
Josef Bacik 已提交
4426
			struct extent_map *hole_em;
Y
Yan Zheng 已提交
4427
			hole_size = last_byte - cur_offset;
J
Josef Bacik 已提交
4428

4429
			trans = btrfs_start_transaction(root, 3);
4430 4431
			if (IS_ERR(trans)) {
				err = PTR_ERR(trans);
J
Josef Bacik 已提交
4432
				break;
4433
			}
4434

J
Josef Bacik 已提交
4435 4436
			err = btrfs_drop_extents(trans, root, inode,
						 cur_offset,
4437
						 cur_offset + hole_size, 1);
4438
			if (err) {
4439
				btrfs_abort_transaction(trans, root, err);
4440
				btrfs_end_transaction(trans, root);
4441
				break;
4442
			}
4443

Y
Yan Zheng 已提交
4444
			err = btrfs_insert_file_extent(trans, root,
L
Li Zefan 已提交
4445
					btrfs_ino(inode), cur_offset, 0,
Y
Yan Zheng 已提交
4446 4447
					0, hole_size, 0, hole_size,
					0, 0, 0);
4448
			if (err) {
4449
				btrfs_abort_transaction(trans, root, err);
4450
				btrfs_end_transaction(trans, root);
4451
				break;
4452
			}
4453

J
Josef Bacik 已提交
4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464
			btrfs_drop_extent_cache(inode, cur_offset,
						cur_offset + hole_size - 1, 0);
			hole_em = alloc_extent_map();
			if (!hole_em) {
				set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
					&BTRFS_I(inode)->runtime_flags);
				goto next;
			}
			hole_em->start = cur_offset;
			hole_em->len = hole_size;
			hole_em->orig_start = cur_offset;
4465

J
Josef Bacik 已提交
4466 4467
			hole_em->block_start = EXTENT_MAP_HOLE;
			hole_em->block_len = 0;
4468
			hole_em->orig_block_len = 0;
J
Josef Bacik 已提交
4469
			hole_em->ram_bytes = hole_size;
J
Josef Bacik 已提交
4470 4471 4472
			hole_em->bdev = root->fs_info->fs_devices->latest_bdev;
			hole_em->compress_type = BTRFS_COMPRESS_NONE;
			hole_em->generation = trans->transid;
4473

J
Josef Bacik 已提交
4474 4475
			while (1) {
				write_lock(&em_tree->lock);
J
Josef Bacik 已提交
4476
				err = add_extent_mapping(em_tree, hole_em, 1);
J
Josef Bacik 已提交
4477 4478 4479 4480 4481 4482 4483 4484 4485
				write_unlock(&em_tree->lock);
				if (err != -EEXIST)
					break;
				btrfs_drop_extent_cache(inode, cur_offset,
							cur_offset +
							hole_size - 1, 0);
			}
			free_extent_map(hole_em);
next:
4486
			btrfs_update_inode(trans, root, inode);
4487
			btrfs_end_transaction(trans, root);
Y
Yan Zheng 已提交
4488 4489
		}
		free_extent_map(em);
4490
		em = NULL;
Y
Yan Zheng 已提交
4491
		cur_offset = last_byte;
4492
		if (cur_offset >= block_end)
Y
Yan Zheng 已提交
4493 4494
			break;
	}
4495

4496
	free_extent_map(em);
4497 4498
	unlock_extent_cached(io_tree, hole_start, block_end - 1, &cached_state,
			     GFP_NOFS);
Y
Yan Zheng 已提交
4499 4500
	return err;
}
C
Chris Mason 已提交
4501

4502
static int btrfs_setsize(struct inode *inode, struct iattr *attr)
4503
{
4504 4505
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
4506
	loff_t oldsize = i_size_read(inode);
4507 4508
	loff_t newsize = attr->ia_size;
	int mask = attr->ia_valid;
4509 4510
	int ret;

4511
	if (newsize == oldsize)
4512 4513
		return 0;

4514 4515 4516 4517 4518 4519 4520 4521 4522
	/*
	 * The regular truncate() case without ATTR_CTIME and ATTR_MTIME is a
	 * special case where we need to update the times despite not having
	 * these flags set.  For all other operations the VFS set these flags
	 * explicitly if it wants a timestamp update.
	 */
	if (newsize != oldsize && (!(mask & (ATTR_CTIME | ATTR_MTIME))))
		inode->i_ctime = inode->i_mtime = current_fs_time(inode->i_sb);

4523 4524 4525
	if (newsize > oldsize) {
		truncate_pagecache(inode, oldsize, newsize);
		ret = btrfs_cont_expand(inode, oldsize, newsize);
4526
		if (ret)
4527 4528
			return ret;

4529 4530 4531 4532 4533 4534 4535
		trans = btrfs_start_transaction(root, 1);
		if (IS_ERR(trans))
			return PTR_ERR(trans);

		i_size_write(inode, newsize);
		btrfs_ordered_update_i_size(inode, i_size_read(inode), NULL);
		ret = btrfs_update_inode(trans, root, inode);
4536
		btrfs_end_transaction(trans, root);
4537
	} else {
4538

4539 4540 4541 4542 4543 4544
		/*
		 * We're truncating a file that used to have good data down to
		 * zero. Make sure it gets into the ordered flush list so that
		 * any new writes get down to disk quickly.
		 */
		if (newsize == 0)
4545 4546
			set_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
				&BTRFS_I(inode)->runtime_flags);
4547

4548 4549 4550 4551 4552 4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570
		/*
		 * 1 for the orphan item we're going to add
		 * 1 for the orphan item deletion.
		 */
		trans = btrfs_start_transaction(root, 2);
		if (IS_ERR(trans))
			return PTR_ERR(trans);

		/*
		 * We need to do this in case we fail at _any_ point during the
		 * actual truncate.  Once we do the truncate_setsize we could
		 * invalidate pages which forces any outstanding ordered io to
		 * be instantly completed which will give us extents that need
		 * to be truncated.  If we fail to get an orphan inode down we
		 * could have left over extents that were never meant to live,
		 * so we need to garuntee from this point on that everything
		 * will be consistent.
		 */
		ret = btrfs_orphan_add(trans, inode);
		btrfs_end_transaction(trans, root);
		if (ret)
			return ret;

4571 4572
		/* we don't support swapfiles, so vmtruncate shouldn't fail */
		truncate_setsize(inode, newsize);
4573 4574 4575 4576 4577 4578

		/* Disable nonlocked read DIO to avoid the end less truncate */
		btrfs_inode_block_unlocked_dio(inode);
		inode_dio_wait(inode);
		btrfs_inode_resume_unlocked_dio(inode);

4579
		ret = btrfs_truncate(inode);
4580 4581
		if (ret && inode->i_nlink)
			btrfs_orphan_del(NULL, inode);
4582 4583
	}

4584
	return ret;
4585 4586
}

Y
Yan Zheng 已提交
4587 4588 4589
static int btrfs_setattr(struct dentry *dentry, struct iattr *attr)
{
	struct inode *inode = dentry->d_inode;
L
Li Zefan 已提交
4590
	struct btrfs_root *root = BTRFS_I(inode)->root;
Y
Yan Zheng 已提交
4591
	int err;
C
Chris Mason 已提交
4592

L
Li Zefan 已提交
4593 4594 4595
	if (btrfs_root_readonly(root))
		return -EROFS;

Y
Yan Zheng 已提交
4596 4597 4598
	err = inode_change_ok(inode, attr);
	if (err)
		return err;
C
Chris Mason 已提交
4599

4600
	if (S_ISREG(inode->i_mode) && (attr->ia_valid & ATTR_SIZE)) {
4601
		err = btrfs_setsize(inode, attr);
4602 4603
		if (err)
			return err;
C
Chris Mason 已提交
4604
	}
Y
Yan Zheng 已提交
4605

C
Christoph Hellwig 已提交
4606 4607
	if (attr->ia_valid) {
		setattr_copy(inode, attr);
4608
		inode_inc_iversion(inode);
4609
		err = btrfs_dirty_inode(inode);
C
Christoph Hellwig 已提交
4610

4611
		if (!err && attr->ia_valid & ATTR_MODE)
C
Christoph Hellwig 已提交
4612 4613
			err = btrfs_acl_chmod(inode);
	}
J
Josef Bacik 已提交
4614

C
Chris Mason 已提交
4615 4616
	return err;
}
4617

A
Al Viro 已提交
4618
void btrfs_evict_inode(struct inode *inode)
C
Chris Mason 已提交
4619 4620 4621
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(inode)->root;
4622
	struct btrfs_block_rsv *rsv, *global_rsv;
4623
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
C
Chris Mason 已提交
4624 4625
	int ret;

4626 4627
	trace_btrfs_inode_evict(inode);

C
Chris Mason 已提交
4628
	truncate_inode_pages(&inode->i_data, 0);
4629
	if (inode->i_nlink && (btrfs_root_refs(&root->root_item) != 0 ||
4630
			       btrfs_is_free_space_inode(inode)))
A
Al Viro 已提交
4631 4632
		goto no_delete;

C
Chris Mason 已提交
4633
	if (is_bad_inode(inode)) {
4634
		btrfs_orphan_del(NULL, inode);
C
Chris Mason 已提交
4635 4636
		goto no_delete;
	}
A
Al Viro 已提交
4637
	/* do we really want it for ->i_nlink > 0 and zero btrfs_root_refs? */
C
Chris Mason 已提交
4638
	btrfs_wait_ordered_range(inode, 0, (u64)-1);
4639

4640
	if (root->fs_info->log_root_recovering) {
4641
		BUG_ON(test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
4642
				 &BTRFS_I(inode)->runtime_flags));
4643 4644 4645
		goto no_delete;
	}

4646 4647 4648 4649 4650
	if (inode->i_nlink > 0) {
		BUG_ON(btrfs_root_refs(&root->root_item) != 0);
		goto no_delete;
	}

4651 4652 4653 4654 4655 4656
	ret = btrfs_commit_inode_delayed_inode(inode);
	if (ret) {
		btrfs_orphan_del(NULL, inode);
		goto no_delete;
	}

4657
	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
4658 4659 4660 4661
	if (!rsv) {
		btrfs_orphan_del(NULL, inode);
		goto no_delete;
	}
J
Josef Bacik 已提交
4662
	rsv->size = min_size;
4663
	rsv->failfast = 1;
4664
	global_rsv = &root->fs_info->global_block_rsv;
4665

4666
	btrfs_i_size_write(inode, 0);
4667

4668
	/*
4669 4670 4671 4672
	 * This is a bit simpler than btrfs_truncate since we've already
	 * reserved our space for our orphan item in the unlink, so we just
	 * need to reserve some slack space in case we add bytes and update
	 * inode item when doing the truncate.
4673
	 */
4674
	while (1) {
M
Miao Xie 已提交
4675 4676
		ret = btrfs_block_rsv_refill(root, rsv, min_size,
					     BTRFS_RESERVE_FLUSH_LIMIT);
4677 4678 4679 4680 4681 4682 4683 4684

		/*
		 * Try and steal from the global reserve since we will
		 * likely not use this space anyway, we want to try as
		 * hard as possible to get this to work.
		 */
		if (ret)
			ret = btrfs_block_rsv_migrate(global_rsv, rsv, min_size);
4685 4686

		if (ret) {
4687 4688 4689
			btrfs_warn(root->fs_info,
				"Could not get space for a delete, will truncate on mount %d",
				ret);
4690 4691 4692
			btrfs_orphan_del(NULL, inode);
			btrfs_free_block_rsv(root, rsv);
			goto no_delete;
4693
		}
4694

4695
		trans = btrfs_join_transaction(root);
4696 4697 4698 4699
		if (IS_ERR(trans)) {
			btrfs_orphan_del(NULL, inode);
			btrfs_free_block_rsv(root, rsv);
			goto no_delete;
4700
		}
4701

4702 4703
		trans->block_rsv = rsv;

4704
		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
4705
		if (ret != -ENOSPC)
4706
			break;
4707

4708
		trans->block_rsv = &root->fs_info->trans_block_rsv;
4709 4710
		btrfs_end_transaction(trans, root);
		trans = NULL;
4711
		btrfs_btree_balance_dirty(root);
4712
	}
4713

4714 4715
	btrfs_free_block_rsv(root, rsv);

4716
	if (ret == 0) {
4717
		trans->block_rsv = root->orphan_block_rsv;
4718 4719 4720
		ret = btrfs_orphan_del(trans, inode);
		BUG_ON(ret);
	}
4721

4722
	trans->block_rsv = &root->fs_info->trans_block_rsv;
4723 4724
	if (!(root == root->fs_info->tree_root ||
	      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
L
Li Zefan 已提交
4725
		btrfs_return_ino(root, btrfs_ino(inode));
4726

4727
	btrfs_end_transaction(trans, root);
4728
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
4729
no_delete:
4730
	btrfs_remove_delayed_node(inode);
4731
	clear_inode(inode);
4732
	return;
C
Chris Mason 已提交
4733 4734 4735 4736 4737 4738 4739 4740 4741 4742 4743 4744 4745 4746
}

/*
 * this returns the key found in the dir entry in the location pointer.
 * If no dir entries were found, location->objectid is 0.
 */
static int btrfs_inode_by_name(struct inode *dir, struct dentry *dentry,
			       struct btrfs_key *location)
{
	const char *name = dentry->d_name.name;
	int namelen = dentry->d_name.len;
	struct btrfs_dir_item *di;
	struct btrfs_path *path;
	struct btrfs_root *root = BTRFS_I(dir)->root;
Y
Yan 已提交
4747
	int ret = 0;
C
Chris Mason 已提交
4748 4749

	path = btrfs_alloc_path();
4750 4751
	if (!path)
		return -ENOMEM;
4752

L
Li Zefan 已提交
4753
	di = btrfs_lookup_dir_item(NULL, root, path, btrfs_ino(dir), name,
C
Chris Mason 已提交
4754
				    namelen, 0);
Y
Yan 已提交
4755 4756
	if (IS_ERR(di))
		ret = PTR_ERR(di);
C
Chris Mason 已提交
4757

4758
	if (IS_ERR_OR_NULL(di))
4759
		goto out_err;
C
Chris Mason 已提交
4760

4761
	btrfs_dir_item_key_to_cpu(path->nodes[0], di, location);
C
Chris Mason 已提交
4762 4763 4764
out:
	btrfs_free_path(path);
	return ret;
4765 4766 4767
out_err:
	location->objectid = 0;
	goto out;
C
Chris Mason 已提交
4768 4769 4770 4771 4772 4773 4774 4775
}

/*
 * when we hit a tree root in a directory, the btrfs part of the inode
 * needs to be changed to reflect the root directory of the tree root.  This
 * is kind of like crossing a mount point.
 */
static int fixup_tree_root_location(struct btrfs_root *root,
4776 4777 4778 4779
				    struct inode *dir,
				    struct dentry *dentry,
				    struct btrfs_key *location,
				    struct btrfs_root **sub_root)
C
Chris Mason 已提交
4780
{
4781 4782 4783 4784 4785 4786
	struct btrfs_path *path;
	struct btrfs_root *new_root;
	struct btrfs_root_ref *ref;
	struct extent_buffer *leaf;
	int ret;
	int err = 0;
C
Chris Mason 已提交
4787

4788 4789 4790 4791 4792
	path = btrfs_alloc_path();
	if (!path) {
		err = -ENOMEM;
		goto out;
	}
C
Chris Mason 已提交
4793

4794 4795 4796 4797 4798 4799 4800 4801 4802
	err = -ENOENT;
	ret = btrfs_find_root_ref(root->fs_info->tree_root, path,
				  BTRFS_I(dir)->root->root_key.objectid,
				  location->objectid);
	if (ret) {
		if (ret < 0)
			err = ret;
		goto out;
	}
C
Chris Mason 已提交
4803

4804 4805
	leaf = path->nodes[0];
	ref = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_root_ref);
L
Li Zefan 已提交
4806
	if (btrfs_root_ref_dirid(leaf, ref) != btrfs_ino(dir) ||
4807 4808
	    btrfs_root_ref_name_len(leaf, ref) != dentry->d_name.len)
		goto out;
C
Chris Mason 已提交
4809

4810 4811 4812 4813 4814 4815
	ret = memcmp_extent_buffer(leaf, dentry->d_name.name,
				   (unsigned long)(ref + 1),
				   dentry->d_name.len);
	if (ret)
		goto out;

4816
	btrfs_release_path(path);
4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836

	new_root = btrfs_read_fs_root_no_name(root->fs_info, location);
	if (IS_ERR(new_root)) {
		err = PTR_ERR(new_root);
		goto out;
	}

	if (btrfs_root_refs(&new_root->root_item) == 0) {
		err = -ENOENT;
		goto out;
	}

	*sub_root = new_root;
	location->objectid = btrfs_root_dirid(&new_root->root_item);
	location->type = BTRFS_INODE_ITEM_KEY;
	location->offset = 0;
	err = 0;
out:
	btrfs_free_path(path);
	return err;
C
Chris Mason 已提交
4837 4838
}

4839 4840 4841 4842
static void inode_tree_add(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_inode *entry;
4843 4844
	struct rb_node **p;
	struct rb_node *parent;
L
Li Zefan 已提交
4845
	u64 ino = btrfs_ino(inode);
4846

A
Al Viro 已提交
4847
	if (inode_unhashed(inode))
4848
		return;
4849 4850
again:
	parent = NULL;
4851
	spin_lock(&root->inode_lock);
4852
	p = &root->inode_tree.rb_node;
4853 4854 4855 4856
	while (*p) {
		parent = *p;
		entry = rb_entry(parent, struct btrfs_inode, rb_node);

L
Li Zefan 已提交
4857
		if (ino < btrfs_ino(&entry->vfs_inode))
4858
			p = &parent->rb_left;
L
Li Zefan 已提交
4859
		else if (ino > btrfs_ino(&entry->vfs_inode))
4860
			p = &parent->rb_right;
4861 4862
		else {
			WARN_ON(!(entry->vfs_inode.i_state &
A
Al Viro 已提交
4863
				  (I_WILL_FREE | I_FREEING)));
4864 4865 4866 4867
			rb_erase(parent, &root->inode_tree);
			RB_CLEAR_NODE(parent);
			spin_unlock(&root->inode_lock);
			goto again;
4868 4869 4870 4871 4872 4873 4874 4875 4876 4877
		}
	}
	rb_link_node(&BTRFS_I(inode)->rb_node, parent, p);
	rb_insert_color(&BTRFS_I(inode)->rb_node, &root->inode_tree);
	spin_unlock(&root->inode_lock);
}

static void inode_tree_del(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
4878
	int empty = 0;
4879

4880
	spin_lock(&root->inode_lock);
4881 4882 4883
	if (!RB_EMPTY_NODE(&BTRFS_I(inode)->rb_node)) {
		rb_erase(&BTRFS_I(inode)->rb_node, &root->inode_tree);
		RB_CLEAR_NODE(&BTRFS_I(inode)->rb_node);
4884
		empty = RB_EMPTY_ROOT(&root->inode_tree);
4885
	}
4886
	spin_unlock(&root->inode_lock);
4887

4888 4889 4890 4891 4892 4893 4894 4895
	/*
	 * Free space cache has inodes in the tree root, but the tree root has a
	 * root_refs of 0, so this could end up dropping the tree root as a
	 * snapshot, so we need the extra !root->fs_info->tree_root check to
	 * make sure we don't drop it.
	 */
	if (empty && btrfs_root_refs(&root->root_item) == 0 &&
	    root != root->fs_info->tree_root) {
4896 4897 4898 4899 4900 4901 4902 4903 4904
		synchronize_srcu(&root->fs_info->subvol_srcu);
		spin_lock(&root->inode_lock);
		empty = RB_EMPTY_ROOT(&root->inode_tree);
		spin_unlock(&root->inode_lock);
		if (empty)
			btrfs_add_dead_root(root);
	}
}

4905
void btrfs_invalidate_inodes(struct btrfs_root *root)
4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916 4917 4918 4919 4920 4921 4922
{
	struct rb_node *node;
	struct rb_node *prev;
	struct btrfs_inode *entry;
	struct inode *inode;
	u64 objectid = 0;

	WARN_ON(btrfs_root_refs(&root->root_item) != 0);

	spin_lock(&root->inode_lock);
again:
	node = root->inode_tree.rb_node;
	prev = NULL;
	while (node) {
		prev = node;
		entry = rb_entry(node, struct btrfs_inode, rb_node);

L
Li Zefan 已提交
4923
		if (objectid < btrfs_ino(&entry->vfs_inode))
4924
			node = node->rb_left;
L
Li Zefan 已提交
4925
		else if (objectid > btrfs_ino(&entry->vfs_inode))
4926 4927 4928 4929 4930 4931 4932
			node = node->rb_right;
		else
			break;
	}
	if (!node) {
		while (prev) {
			entry = rb_entry(prev, struct btrfs_inode, rb_node);
L
Li Zefan 已提交
4933
			if (objectid <= btrfs_ino(&entry->vfs_inode)) {
4934 4935 4936 4937 4938 4939 4940 4941
				node = prev;
				break;
			}
			prev = rb_next(prev);
		}
	}
	while (node) {
		entry = rb_entry(node, struct btrfs_inode, rb_node);
L
Li Zefan 已提交
4942
		objectid = btrfs_ino(&entry->vfs_inode) + 1;
4943 4944 4945 4946 4947 4948
		inode = igrab(&entry->vfs_inode);
		if (inode) {
			spin_unlock(&root->inode_lock);
			if (atomic_read(&inode->i_count) > 1)
				d_prune_aliases(inode);
			/*
4949
			 * btrfs_drop_inode will have it removed from
4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964
			 * the inode cache when its usage count
			 * hits zero.
			 */
			iput(inode);
			cond_resched();
			spin_lock(&root->inode_lock);
			goto again;
		}

		if (cond_resched_lock(&root->inode_lock))
			goto again;

		node = rb_next(node);
	}
	spin_unlock(&root->inode_lock);
4965 4966
}

4967 4968 4969 4970 4971
static int btrfs_init_locked_inode(struct inode *inode, void *p)
{
	struct btrfs_iget_args *args = p;
	inode->i_ino = args->ino;
	BTRFS_I(inode)->root = args->root;
C
Chris Mason 已提交
4972 4973 4974 4975 4976 4977
	return 0;
}

static int btrfs_find_actor(struct inode *inode, void *opaque)
{
	struct btrfs_iget_args *args = opaque;
L
Li Zefan 已提交
4978
	return args->ino == btrfs_ino(inode) &&
C
Chris Mason 已提交
4979
		args->root == BTRFS_I(inode)->root;
C
Chris Mason 已提交
4980 4981
}

4982 4983 4984
static struct inode *btrfs_iget_locked(struct super_block *s,
				       u64 objectid,
				       struct btrfs_root *root)
C
Chris Mason 已提交
4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996
{
	struct inode *inode;
	struct btrfs_iget_args args;
	args.ino = objectid;
	args.root = root;

	inode = iget5_locked(s, objectid, btrfs_find_actor,
			     btrfs_init_locked_inode,
			     (void *)&args);
	return inode;
}

B
Balaji Rao 已提交
4997 4998 4999 5000
/* Get an inode object given its location and corresponding root.
 * Returns in *is_new if the inode was read from disk
 */
struct inode *btrfs_iget(struct super_block *s, struct btrfs_key *location,
5001
			 struct btrfs_root *root, int *new)
B
Balaji Rao 已提交
5002 5003 5004 5005 5006
{
	struct inode *inode;

	inode = btrfs_iget_locked(s, location->objectid, root);
	if (!inode)
5007
		return ERR_PTR(-ENOMEM);
B
Balaji Rao 已提交
5008 5009 5010 5011 5012

	if (inode->i_state & I_NEW) {
		BTRFS_I(inode)->root = root;
		memcpy(&BTRFS_I(inode)->location, location, sizeof(*location));
		btrfs_read_locked_inode(inode);
5013 5014 5015 5016 5017 5018
		if (!is_bad_inode(inode)) {
			inode_tree_add(inode);
			unlock_new_inode(inode);
			if (new)
				*new = 1;
		} else {
5019 5020 5021
			unlock_new_inode(inode);
			iput(inode);
			inode = ERR_PTR(-ESTALE);
5022 5023 5024
		}
	}

B
Balaji Rao 已提交
5025 5026 5027
	return inode;
}

5028 5029 5030 5031 5032 5033 5034 5035 5036 5037 5038
static struct inode *new_simple_dir(struct super_block *s,
				    struct btrfs_key *key,
				    struct btrfs_root *root)
{
	struct inode *inode = new_inode(s);

	if (!inode)
		return ERR_PTR(-ENOMEM);

	BTRFS_I(inode)->root = root;
	memcpy(&BTRFS_I(inode)->location, key, sizeof(*key));
5039
	set_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags);
5040 5041

	inode->i_ino = BTRFS_EMPTY_SUBVOL_DIR_OBJECTID;
L
Li Zefan 已提交
5042
	inode->i_op = &btrfs_dir_ro_inode_operations;
5043 5044 5045 5046 5047 5048 5049
	inode->i_fop = &simple_dir_operations;
	inode->i_mode = S_IFDIR | S_IRUGO | S_IWUSR | S_IXUGO;
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;

	return inode;
}

5050
struct inode *btrfs_lookup_dentry(struct inode *dir, struct dentry *dentry)
C
Chris Mason 已提交
5051
{
C
Chris Mason 已提交
5052
	struct inode *inode;
5053
	struct btrfs_root *root = BTRFS_I(dir)->root;
C
Chris Mason 已提交
5054 5055
	struct btrfs_root *sub_root = root;
	struct btrfs_key location;
5056
	int index;
5057
	int ret = 0;
C
Chris Mason 已提交
5058 5059 5060

	if (dentry->d_name.len > BTRFS_NAME_LEN)
		return ERR_PTR(-ENAMETOOLONG);
5061

J
Jeff Layton 已提交
5062
	ret = btrfs_inode_by_name(dir, dentry, &location);
C
Chris Mason 已提交
5063 5064
	if (ret < 0)
		return ERR_PTR(ret);
5065

5066 5067 5068 5069
	if (location.objectid == 0)
		return NULL;

	if (location.type == BTRFS_INODE_ITEM_KEY) {
5070
		inode = btrfs_iget(dir->i_sb, &location, root, NULL);
5071 5072 5073 5074 5075
		return inode;
	}

	BUG_ON(location.type != BTRFS_ROOT_ITEM_KEY);

5076
	index = srcu_read_lock(&root->fs_info->subvol_srcu);
5077 5078 5079 5080 5081 5082 5083 5084
	ret = fixup_tree_root_location(root, dir, dentry,
				       &location, &sub_root);
	if (ret < 0) {
		if (ret != -ENOENT)
			inode = ERR_PTR(ret);
		else
			inode = new_simple_dir(dir->i_sb, &location, sub_root);
	} else {
5085
		inode = btrfs_iget(dir->i_sb, &location, sub_root, NULL);
C
Chris Mason 已提交
5086
	}
5087 5088
	srcu_read_unlock(&root->fs_info->subvol_srcu, index);

5089
	if (!IS_ERR(inode) && root != sub_root) {
5090 5091
		down_read(&root->fs_info->cleanup_work_sem);
		if (!(inode->i_sb->s_flags & MS_RDONLY))
5092
			ret = btrfs_orphan_cleanup(sub_root);
5093
		up_read(&root->fs_info->cleanup_work_sem);
5094 5095
		if (ret)
			inode = ERR_PTR(ret);
5096 5097
	}

5098 5099 5100
	return inode;
}

N
Nick Piggin 已提交
5101
static int btrfs_dentry_delete(const struct dentry *dentry)
5102 5103
{
	struct btrfs_root *root;
L
Li Zefan 已提交
5104
	struct inode *inode = dentry->d_inode;
5105

L
Li Zefan 已提交
5106 5107
	if (!inode && !IS_ROOT(dentry))
		inode = dentry->d_parent->d_inode;
5108

L
Li Zefan 已提交
5109 5110
	if (inode) {
		root = BTRFS_I(inode)->root;
5111 5112
		if (btrfs_root_refs(&root->root_item) == 0)
			return 1;
L
Li Zefan 已提交
5113 5114 5115

		if (btrfs_ino(inode) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
			return 1;
5116
	}
5117 5118 5119
	return 0;
}

5120 5121 5122 5123 5124 5125
static void btrfs_dentry_release(struct dentry *dentry)
{
	if (dentry->d_fsdata)
		kfree(dentry->d_fsdata);
}

5126
static struct dentry *btrfs_lookup(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
5127
				   unsigned int flags)
5128
{
5129 5130 5131 5132
	struct dentry *ret;

	ret = d_splice_alias(btrfs_lookup_dentry(dir, dentry), dentry);
	return ret;
C
Chris Mason 已提交
5133 5134
}

5135
unsigned char btrfs_filetype_table[] = {
C
Chris Mason 已提交
5136 5137 5138
	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
};

5139 5140
static int btrfs_real_readdir(struct file *filp, void *dirent,
			      filldir_t filldir)
C
Chris Mason 已提交
5141
{
A
Al Viro 已提交
5142
	struct inode *inode = file_inode(filp);
C
Chris Mason 已提交
5143 5144 5145 5146
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_item *item;
	struct btrfs_dir_item *di;
	struct btrfs_key key;
5147
	struct btrfs_key found_key;
C
Chris Mason 已提交
5148
	struct btrfs_path *path;
5149 5150
	struct list_head ins_list;
	struct list_head del_list;
C
Chris Mason 已提交
5151
	int ret;
5152
	struct extent_buffer *leaf;
C
Chris Mason 已提交
5153 5154 5155 5156 5157 5158 5159
	int slot;
	unsigned char d_type;
	int over = 0;
	u32 di_cur;
	u32 di_total;
	u32 di_len;
	int key_type = BTRFS_DIR_INDEX_KEY;
5160 5161 5162
	char tmp_name[32];
	char *name_ptr;
	int name_len;
5163
	int is_curr = 0;	/* filp->f_pos points to the current index? */
C
Chris Mason 已提交
5164 5165 5166 5167

	/* FIXME, use a real flag for deciding about the key type */
	if (root->fs_info->tree_root == root)
		key_type = BTRFS_DIR_ITEM_KEY;
5168

5169 5170
	/* special case for "." */
	if (filp->f_pos == 0) {
5171 5172
		over = filldir(dirent, ".", 1,
			       filp->f_pos, btrfs_ino(inode), DT_DIR);
5173 5174 5175 5176 5177 5178
		if (over)
			return 0;
		filp->f_pos = 1;
	}
	/* special case for .., just use the back ref */
	if (filp->f_pos == 1) {
5179
		u64 pino = parent_ino(filp->f_path.dentry);
5180
		over = filldir(dirent, "..", 2,
5181
			       filp->f_pos, pino, DT_DIR);
5182
		if (over)
5183
			return 0;
5184 5185
		filp->f_pos = 2;
	}
5186
	path = btrfs_alloc_path();
5187 5188
	if (!path)
		return -ENOMEM;
C
Chris Mason 已提交
5189

J
Josef Bacik 已提交
5190
	path->reada = 1;
5191

5192 5193 5194 5195 5196 5197
	if (key_type == BTRFS_DIR_INDEX_KEY) {
		INIT_LIST_HEAD(&ins_list);
		INIT_LIST_HEAD(&del_list);
		btrfs_get_delayed_items(inode, &ins_list, &del_list);
	}

C
Chris Mason 已提交
5198 5199
	btrfs_set_key_type(&key, key_type);
	key.offset = filp->f_pos;
L
Li Zefan 已提交
5200
	key.objectid = btrfs_ino(inode);
5201

C
Chris Mason 已提交
5202 5203 5204
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto err;
5205 5206

	while (1) {
5207
		leaf = path->nodes[0];
C
Chris Mason 已提交
5208
		slot = path->slots[0];
5209 5210 5211 5212 5213 5214 5215
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto err;
			else if (ret > 0)
				break;
			continue;
C
Chris Mason 已提交
5216
		}
5217

5218 5219 5220 5221
		item = btrfs_item_nr(leaf, slot);
		btrfs_item_key_to_cpu(leaf, &found_key, slot);

		if (found_key.objectid != key.objectid)
C
Chris Mason 已提交
5222
			break;
5223
		if (btrfs_key_type(&found_key) != key_type)
C
Chris Mason 已提交
5224
			break;
5225
		if (found_key.offset < filp->f_pos)
5226
			goto next;
5227 5228 5229 5230
		if (key_type == BTRFS_DIR_INDEX_KEY &&
		    btrfs_should_delete_dir_index(&del_list,
						  found_key.offset))
			goto next;
5231 5232

		filp->f_pos = found_key.offset;
5233
		is_curr = 1;
5234

C
Chris Mason 已提交
5235 5236
		di = btrfs_item_ptr(leaf, slot, struct btrfs_dir_item);
		di_cur = 0;
5237
		di_total = btrfs_item_size(leaf, item);
5238 5239

		while (di_cur < di_total) {
5240 5241
			struct btrfs_key location;

5242 5243 5244
			if (verify_dir_item(root, leaf, di))
				break;

5245
			name_len = btrfs_dir_name_len(leaf, di);
5246
			if (name_len <= sizeof(tmp_name)) {
5247 5248 5249
				name_ptr = tmp_name;
			} else {
				name_ptr = kmalloc(name_len, GFP_NOFS);
5250 5251 5252 5253
				if (!name_ptr) {
					ret = -ENOMEM;
					goto err;
				}
5254 5255 5256 5257 5258 5259
			}
			read_extent_buffer(leaf, name_ptr,
					   (unsigned long)(di + 1), name_len);

			d_type = btrfs_filetype_table[btrfs_dir_type(leaf, di)];
			btrfs_dir_item_key_to_cpu(leaf, di, &location);
5260

5261

5262
			/* is this a reference to our own snapshot? If so
A
Arne Jansen 已提交
5263 5264 5265 5266 5267 5268 5269
			 * skip it.
			 *
			 * In contrast to old kernels, we insert the snapshot's
			 * dir item and dir index after it has been created, so
			 * we won't find a reference to our own snapshot. We
			 * still keep the following code for backward
			 * compatibility.
5270 5271 5272 5273 5274 5275
			 */
			if (location.type == BTRFS_ROOT_ITEM_KEY &&
			    location.objectid == root->root_key.objectid) {
				over = 0;
				goto skip;
			}
5276
			over = filldir(dirent, name_ptr, name_len,
5277
				       found_key.offset, location.objectid,
C
Chris Mason 已提交
5278
				       d_type);
5279

5280
skip:
5281 5282 5283
			if (name_ptr != tmp_name)
				kfree(name_ptr);

C
Chris Mason 已提交
5284 5285
			if (over)
				goto nopos;
J
Josef Bacik 已提交
5286
			di_len = btrfs_dir_name_len(leaf, di) +
5287
				 btrfs_dir_data_len(leaf, di) + sizeof(*di);
C
Chris Mason 已提交
5288 5289 5290
			di_cur += di_len;
			di = (struct btrfs_dir_item *)((char *)di + di_len);
		}
5291 5292
next:
		path->slots[0]++;
C
Chris Mason 已提交
5293
	}
5294

5295 5296 5297 5298 5299 5300 5301 5302 5303
	if (key_type == BTRFS_DIR_INDEX_KEY) {
		if (is_curr)
			filp->f_pos++;
		ret = btrfs_readdir_delayed_dir_index(filp, dirent, filldir,
						      &ins_list);
		if (ret)
			goto nopos;
	}

5304
	/* Reached end of directory/root. Bump pos past the last item. */
5305
	if (key_type == BTRFS_DIR_INDEX_KEY)
5306 5307 5308 5309 5310
		/*
		 * 32-bit glibc will use getdents64, but then strtol -
		 * so the last number we can serve is this.
		 */
		filp->f_pos = 0x7fffffff;
5311 5312
	else
		filp->f_pos++;
C
Chris Mason 已提交
5313 5314 5315
nopos:
	ret = 0;
err:
5316 5317
	if (key_type == BTRFS_DIR_INDEX_KEY)
		btrfs_put_delayed_items(&ins_list, &del_list);
C
Chris Mason 已提交
5318 5319 5320 5321
	btrfs_free_path(path);
	return ret;
}

5322
int btrfs_write_inode(struct inode *inode, struct writeback_control *wbc)
C
Chris Mason 已提交
5323 5324 5325 5326
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
	int ret = 0;
5327
	bool nolock = false;
C
Chris Mason 已提交
5328

5329
	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
5330 5331
		return 0;

5332
	if (btrfs_fs_closing(root->fs_info) && btrfs_is_free_space_inode(inode))
5333
		nolock = true;
5334

5335
	if (wbc->sync_mode == WB_SYNC_ALL) {
5336
		if (nolock)
5337
			trans = btrfs_join_transaction_nolock(root);
5338
		else
5339
			trans = btrfs_join_transaction(root);
5340 5341
		if (IS_ERR(trans))
			return PTR_ERR(trans);
5342
		ret = btrfs_commit_transaction(trans, root);
C
Chris Mason 已提交
5343 5344 5345 5346 5347
	}
	return ret;
}

/*
5348
 * This is somewhat expensive, updating the tree every time the
C
Chris Mason 已提交
5349 5350 5351 5352
 * inode changes.  But, it is most likely to find the inode in cache.
 * FIXME, needs more benchmarking...there are no reasons other than performance
 * to keep or drop this code.
 */
5353
static int btrfs_dirty_inode(struct inode *inode)
C
Chris Mason 已提交
5354 5355 5356
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
5357 5358
	int ret;

5359
	if (test_bit(BTRFS_INODE_DUMMY, &BTRFS_I(inode)->runtime_flags))
5360
		return 0;
C
Chris Mason 已提交
5361

5362
	trans = btrfs_join_transaction(root);
5363 5364
	if (IS_ERR(trans))
		return PTR_ERR(trans);
5365 5366

	ret = btrfs_update_inode(trans, root, inode);
5367 5368 5369 5370
	if (ret && ret == -ENOSPC) {
		/* whoops, lets try again with the full transaction */
		btrfs_end_transaction(trans, root);
		trans = btrfs_start_transaction(root, 1);
5371 5372
		if (IS_ERR(trans))
			return PTR_ERR(trans);
5373

5374 5375
		ret = btrfs_update_inode(trans, root, inode);
	}
C
Chris Mason 已提交
5376
	btrfs_end_transaction(trans, root);
5377 5378
	if (BTRFS_I(inode)->delayed_node)
		btrfs_balance_delayed_items(root);
5379 5380 5381 5382 5383 5384 5385 5386

	return ret;
}

/*
 * This is a copy of file_update_time.  We need this so we can return error on
 * ENOSPC for updating the inode in the case of file write and mmap writes.
 */
5387 5388
static int btrfs_update_time(struct inode *inode, struct timespec *now,
			     int flags)
5389
{
5390 5391 5392 5393 5394
	struct btrfs_root *root = BTRFS_I(inode)->root;

	if (btrfs_root_readonly(root))
		return -EROFS;

5395
	if (flags & S_VERSION)
5396
		inode_inc_iversion(inode);
5397 5398 5399 5400 5401 5402 5403
	if (flags & S_CTIME)
		inode->i_ctime = *now;
	if (flags & S_MTIME)
		inode->i_mtime = *now;
	if (flags & S_ATIME)
		inode->i_atime = *now;
	return btrfs_dirty_inode(inode);
C
Chris Mason 已提交
5404 5405
}

C
Chris Mason 已提交
5406 5407 5408 5409 5410
/*
 * find the highest existing sequence number in a directory
 * and then set the in-memory index_cnt variable to reflect
 * free sequence numbers
 */
5411 5412 5413 5414 5415 5416 5417 5418
static int btrfs_set_inode_index_count(struct inode *inode)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key key, found_key;
	struct btrfs_path *path;
	struct extent_buffer *leaf;
	int ret;

L
Li Zefan 已提交
5419
	key.objectid = btrfs_ino(inode);
5420 5421 5422 5423 5424 5425 5426 5427 5428 5429 5430 5431 5432 5433 5434 5435 5436 5437 5438 5439 5440 5441 5442 5443 5444 5445 5446 5447 5448 5449 5450
	btrfs_set_key_type(&key, BTRFS_DIR_INDEX_KEY);
	key.offset = (u64)-1;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		goto out;
	/* FIXME: we should be able to handle this */
	if (ret == 0)
		goto out;
	ret = 0;

	/*
	 * MAGIC NUMBER EXPLANATION:
	 * since we search a directory based on f_pos we have to start at 2
	 * since '.' and '..' have f_pos of 0 and 1 respectively, so everybody
	 * else has to start at 2
	 */
	if (path->slots[0] == 0) {
		BTRFS_I(inode)->index_cnt = 2;
		goto out;
	}

	path->slots[0]--;

	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);

L
Li Zefan 已提交
5451
	if (found_key.objectid != btrfs_ino(inode) ||
5452 5453 5454 5455 5456 5457 5458 5459 5460 5461 5462
	    btrfs_key_type(&found_key) != BTRFS_DIR_INDEX_KEY) {
		BTRFS_I(inode)->index_cnt = 2;
		goto out;
	}

	BTRFS_I(inode)->index_cnt = found_key.offset + 1;
out:
	btrfs_free_path(path);
	return ret;
}

C
Chris Mason 已提交
5463 5464 5465 5466
/*
 * helper to find a free sequence number in a given directory.  This current
 * code is very simple, later versions will do smarter things in the btree
 */
5467
int btrfs_set_inode_index(struct inode *dir, u64 *index)
5468 5469 5470 5471
{
	int ret = 0;

	if (BTRFS_I(dir)->index_cnt == (u64)-1) {
5472 5473 5474 5475 5476 5477
		ret = btrfs_inode_delayed_dir_index_count(dir);
		if (ret) {
			ret = btrfs_set_inode_index_count(dir);
			if (ret)
				return ret;
		}
5478 5479
	}

5480
	*index = BTRFS_I(dir)->index_cnt;
5481 5482 5483 5484 5485
	BTRFS_I(dir)->index_cnt++;

	return ret;
}

C
Chris Mason 已提交
5486 5487
static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
5488
				     struct inode *dir,
5489
				     const char *name, int name_len,
A
Al Viro 已提交
5490 5491
				     u64 ref_objectid, u64 objectid,
				     umode_t mode, u64 *index)
C
Chris Mason 已提交
5492 5493
{
	struct inode *inode;
5494
	struct btrfs_inode_item *inode_item;
C
Chris Mason 已提交
5495
	struct btrfs_key *location;
5496
	struct btrfs_path *path;
5497 5498 5499 5500
	struct btrfs_inode_ref *ref;
	struct btrfs_key key[2];
	u32 sizes[2];
	unsigned long ptr;
C
Chris Mason 已提交
5501 5502 5503
	int ret;
	int owner;

5504
	path = btrfs_alloc_path();
5505 5506
	if (!path)
		return ERR_PTR(-ENOMEM);
5507

C
Chris Mason 已提交
5508
	inode = new_inode(root->fs_info->sb);
5509 5510
	if (!inode) {
		btrfs_free_path(path);
C
Chris Mason 已提交
5511
		return ERR_PTR(-ENOMEM);
5512
	}
C
Chris Mason 已提交
5513

5514 5515 5516 5517 5518 5519
	/*
	 * we have to initialize this early, so we can reclaim the inode
	 * number if we fail afterwards in this function.
	 */
	inode->i_ino = objectid;

5520
	if (dir) {
5521 5522
		trace_btrfs_inode_request(dir);

5523
		ret = btrfs_set_inode_index(dir, index);
5524
		if (ret) {
5525
			btrfs_free_path(path);
5526
			iput(inode);
5527
			return ERR_PTR(ret);
5528
		}
5529 5530 5531 5532 5533 5534 5535
	}
	/*
	 * index_cnt is ignored for everything but a dir,
	 * btrfs_get_inode_index_count has an explanation for the magic
	 * number
	 */
	BTRFS_I(inode)->index_cnt = 2;
C
Chris Mason 已提交
5536
	BTRFS_I(inode)->root = root;
5537
	BTRFS_I(inode)->generation = trans->transid;
5538
	inode->i_generation = BTRFS_I(inode)->generation;
5539

J
Josef Bacik 已提交
5540 5541 5542 5543 5544 5545 5546 5547
	/*
	 * We could have gotten an inode number from somebody who was fsynced
	 * and then removed in this same transaction, so let's just set full
	 * sync since it will be a full sync anyway and this will blow away the
	 * old info in the log.
	 */
	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);

5548
	if (S_ISDIR(mode))
C
Chris Mason 已提交
5549 5550 5551
		owner = 0;
	else
		owner = 1;
5552 5553 5554 5555 5556

	key[0].objectid = objectid;
	btrfs_set_key_type(&key[0], BTRFS_INODE_ITEM_KEY);
	key[0].offset = 0;

M
Mark Fasheh 已提交
5557 5558 5559 5560 5561 5562
	/*
	 * Start new inodes with an inode_ref. This is slightly more
	 * efficient for small numbers of hard links since they will
	 * be packed into one item. Extended refs will kick in if we
	 * add more hard links than can fit in the ref item.
	 */
5563 5564 5565 5566 5567 5568 5569
	key[1].objectid = objectid;
	btrfs_set_key_type(&key[1], BTRFS_INODE_REF_KEY);
	key[1].offset = ref_objectid;

	sizes[0] = sizeof(struct btrfs_inode_item);
	sizes[1] = name_len + sizeof(*ref);

5570
	path->leave_spinning = 1;
5571 5572
	ret = btrfs_insert_empty_items(trans, root, path, key, sizes, 2);
	if (ret != 0)
5573 5574
		goto fail;

5575
	inode_init_owner(inode, dir, mode);
5576
	inode_set_bytes(inode, 0);
C
Chris Mason 已提交
5577
	inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME;
5578 5579
	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_inode_item);
5580 5581
	memset_extent_buffer(path->nodes[0], 0, (unsigned long)inode_item,
			     sizeof(*inode_item));
5582
	fill_inode_item(trans, path->nodes[0], inode_item, inode);
5583 5584 5585 5586

	ref = btrfs_item_ptr(path->nodes[0], path->slots[0] + 1,
			     struct btrfs_inode_ref);
	btrfs_set_inode_ref_name_len(path->nodes[0], ref, name_len);
5587
	btrfs_set_inode_ref_index(path->nodes[0], ref, *index);
5588 5589 5590
	ptr = (unsigned long)(ref + 1);
	write_extent_buffer(path->nodes[0], name, ptr, name_len);

5591 5592 5593
	btrfs_mark_buffer_dirty(path->nodes[0]);
	btrfs_free_path(path);

C
Chris Mason 已提交
5594 5595 5596 5597 5598
	location = &BTRFS_I(inode)->location;
	location->objectid = objectid;
	location->offset = 0;
	btrfs_set_key_type(location, BTRFS_INODE_ITEM_KEY);

5599 5600
	btrfs_inherit_iflags(inode, dir);

5601
	if (S_ISREG(mode)) {
5602 5603
		if (btrfs_test_opt(root, NODATASUM))
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATASUM;
L
Liu Bo 已提交
5604
		if (btrfs_test_opt(root, NODATACOW))
5605 5606
			BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW |
				BTRFS_INODE_NODATASUM;
5607 5608
	}

C
Chris Mason 已提交
5609
	insert_inode_hash(inode);
5610
	inode_tree_add(inode);
5611 5612

	trace_btrfs_inode_new(inode);
5613
	btrfs_set_inode_last_trans(trans, inode);
5614

5615 5616
	btrfs_update_root_times(trans, root);

C
Chris Mason 已提交
5617
	return inode;
5618
fail:
5619 5620
	if (dir)
		BTRFS_I(dir)->index_cnt--;
5621
	btrfs_free_path(path);
5622
	iput(inode);
5623
	return ERR_PTR(ret);
C
Chris Mason 已提交
5624 5625 5626 5627 5628 5629 5630
}

static inline u8 btrfs_inode_type(struct inode *inode)
{
	return btrfs_type_by_mode[(inode->i_mode & S_IFMT) >> S_SHIFT];
}

C
Chris Mason 已提交
5631 5632 5633 5634 5635 5636
/*
 * utility function to add 'inode' into 'parent_inode' with
 * a give name and a given sequence number.
 * if 'add_backref' is true, also insert a backref from the
 * inode to the parent directory.
 */
5637 5638 5639
int btrfs_add_link(struct btrfs_trans_handle *trans,
		   struct inode *parent_inode, struct inode *inode,
		   const char *name, int name_len, int add_backref, u64 index)
C
Chris Mason 已提交
5640
{
5641
	int ret = 0;
C
Chris Mason 已提交
5642
	struct btrfs_key key;
5643
	struct btrfs_root *root = BTRFS_I(parent_inode)->root;
L
Li Zefan 已提交
5644 5645
	u64 ino = btrfs_ino(inode);
	u64 parent_ino = btrfs_ino(parent_inode);
5646

L
Li Zefan 已提交
5647
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
5648 5649
		memcpy(&key, &BTRFS_I(inode)->root->root_key, sizeof(key));
	} else {
L
Li Zefan 已提交
5650
		key.objectid = ino;
5651 5652 5653 5654
		btrfs_set_key_type(&key, BTRFS_INODE_ITEM_KEY);
		key.offset = 0;
	}

L
Li Zefan 已提交
5655
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
5656 5657
		ret = btrfs_add_root_ref(trans, root->fs_info->tree_root,
					 key.objectid, root->root_key.objectid,
L
Li Zefan 已提交
5658
					 parent_ino, index, name, name_len);
5659
	} else if (add_backref) {
L
Li Zefan 已提交
5660 5661
		ret = btrfs_insert_inode_ref(trans, root, name, name_len, ino,
					     parent_ino, index);
5662
	}
C
Chris Mason 已提交
5663

5664 5665 5666
	/* Nothing to clean up yet */
	if (ret)
		return ret;
5667

5668 5669 5670
	ret = btrfs_insert_dir_item(trans, root, name, name_len,
				    parent_inode, &key,
				    btrfs_inode_type(inode), index);
C
Chris Mason 已提交
5671
	if (ret == -EEXIST || ret == -EOVERFLOW)
5672 5673 5674 5675
		goto fail_dir_item;
	else if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		return ret;
C
Chris Mason 已提交
5676
	}
5677 5678 5679

	btrfs_i_size_write(parent_inode, parent_inode->i_size +
			   name_len * 2);
5680
	inode_inc_iversion(parent_inode);
5681 5682 5683 5684
	parent_inode->i_mtime = parent_inode->i_ctime = CURRENT_TIME;
	ret = btrfs_update_inode(trans, root, parent_inode);
	if (ret)
		btrfs_abort_transaction(trans, root, ret);
C
Chris Mason 已提交
5685
	return ret;
5686 5687 5688 5689 5690 5691 5692 5693 5694 5695 5696 5697 5698 5699 5700 5701 5702

fail_dir_item:
	if (unlikely(ino == BTRFS_FIRST_FREE_OBJECTID)) {
		u64 local_index;
		int err;
		err = btrfs_del_root_ref(trans, root->fs_info->tree_root,
				 key.objectid, root->root_key.objectid,
				 parent_ino, &local_index, name, name_len);

	} else if (add_backref) {
		u64 local_index;
		int err;

		err = btrfs_del_inode_ref(trans, root, name, name_len,
					  ino, parent_ino, &local_index);
	}
	return ret;
C
Chris Mason 已提交
5703 5704 5705
}

static int btrfs_add_nondir(struct btrfs_trans_handle *trans,
5706 5707
			    struct inode *dir, struct dentry *dentry,
			    struct inode *inode, int backref, u64 index)
C
Chris Mason 已提交
5708
{
5709 5710 5711
	int err = btrfs_add_link(trans, dir, inode,
				 dentry->d_name.name, dentry->d_name.len,
				 backref, index);
C
Chris Mason 已提交
5712 5713 5714 5715 5716
	if (err > 0)
		err = -EEXIST;
	return err;
}

J
Josef Bacik 已提交
5717
static int btrfs_mknod(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
5718
			umode_t mode, dev_t rdev)
J
Josef Bacik 已提交
5719 5720 5721
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
5722
	struct inode *inode = NULL;
J
Josef Bacik 已提交
5723 5724 5725
	int err;
	int drop_inode = 0;
	u64 objectid;
5726
	u64 index = 0;
J
Josef Bacik 已提交
5727 5728 5729 5730

	if (!new_valid_dev(rdev))
		return -EINVAL;

J
Josef Bacik 已提交
5731 5732 5733 5734 5735
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
5736 5737 5738
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
5739

5740 5741 5742 5743
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

5744
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5745
				dentry->d_name.len, btrfs_ino(dir), objectid,
5746
				mode, &index);
5747 5748
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
J
Josef Bacik 已提交
5749
		goto out_unlock;
5750
	}
J
Josef Bacik 已提交
5751

5752
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
5753 5754 5755 5756 5757
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

5758 5759 5760 5761 5762 5763 5764 5765
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/

	inode->i_op = &btrfs_special_inode_operations;
5766
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
J
Josef Bacik 已提交
5767 5768 5769 5770
	if (err)
		drop_inode = 1;
	else {
		init_special_inode(inode, inode->i_mode, rdev);
5771
		btrfs_update_inode(trans, root, inode);
5772
		d_instantiate(dentry, inode);
J
Josef Bacik 已提交
5773 5774
	}
out_unlock:
5775
	btrfs_end_transaction(trans, root);
5776
	btrfs_btree_balance_dirty(root);
J
Josef Bacik 已提交
5777 5778 5779 5780 5781 5782 5783
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
	return err;
}

C
Chris Mason 已提交
5784
static int btrfs_create(struct inode *dir, struct dentry *dentry,
A
Al Viro 已提交
5785
			umode_t mode, bool excl)
C
Chris Mason 已提交
5786 5787 5788
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
5789
	struct inode *inode = NULL;
5790
	int drop_inode_on_err = 0;
5791
	int err;
C
Chris Mason 已提交
5792
	u64 objectid;
5793
	u64 index = 0;
C
Chris Mason 已提交
5794

J
Josef Bacik 已提交
5795 5796 5797 5798 5799
	/*
	 * 2 for inode item and ref
	 * 2 for dir items
	 * 1 for xattr if selinux is on
	 */
5800 5801 5802
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
J
Josef Bacik 已提交
5803

5804 5805 5806 5807
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

5808
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5809
				dentry->d_name.len, btrfs_ino(dir), objectid,
5810
				mode, &index);
5811 5812
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
C
Chris Mason 已提交
5813
		goto out_unlock;
5814
	}
5815
	drop_inode_on_err = 1;
C
Chris Mason 已提交
5816

5817
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
5818
	if (err)
J
Josef Bacik 已提交
5819 5820
		goto out_unlock;

5821 5822 5823 5824
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_unlock;

5825 5826 5827 5828 5829 5830 5831 5832 5833
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/
	inode->i_fop = &btrfs_file_operations;
	inode->i_op = &btrfs_file_inode_operations;

5834
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
5835
	if (err)
5836 5837 5838 5839 5840 5841 5842
		goto out_unlock;

	inode->i_mapping->a_ops = &btrfs_aops;
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
	BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
	d_instantiate(dentry, inode);

C
Chris Mason 已提交
5843
out_unlock:
5844
	btrfs_end_transaction(trans, root);
5845
	if (err && drop_inode_on_err) {
C
Chris Mason 已提交
5846 5847 5848
		inode_dec_link_count(inode);
		iput(inode);
	}
5849
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
5850 5851 5852 5853 5854 5855 5856 5857 5858
	return err;
}

static int btrfs_link(struct dentry *old_dentry, struct inode *dir,
		      struct dentry *dentry)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct inode *inode = old_dentry->d_inode;
5859
	u64 index;
C
Chris Mason 已提交
5860 5861 5862
	int err;
	int drop_inode = 0;

5863 5864
	/* do not allow sys_link's with other subvols of the same device */
	if (root->objectid != BTRFS_I(inode)->root->objectid)
5865
		return -EXDEV;
5866

M
Mark Fasheh 已提交
5867
	if (inode->i_nlink >= BTRFS_LINK_MAX)
5868
		return -EMLINK;
5869

5870
	err = btrfs_set_inode_index(dir, &index);
5871 5872 5873
	if (err)
		goto fail;

5874
	/*
M
Miao Xie 已提交
5875
	 * 2 items for inode and inode ref
5876
	 * 2 items for dir items
M
Miao Xie 已提交
5877
	 * 1 item for parent inode
5878
	 */
M
Miao Xie 已提交
5879
	trans = btrfs_start_transaction(root, 5);
5880 5881 5882 5883
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto fail;
	}
5884

5885
	btrfs_inc_nlink(inode);
5886
	inode_inc_iversion(inode);
5887
	inode->i_ctime = CURRENT_TIME;
A
Al Viro 已提交
5888
	ihold(inode);
5889
	set_bit(BTRFS_INODE_COPY_EVERYTHING, &BTRFS_I(inode)->runtime_flags);
5890

5891
	err = btrfs_add_nondir(trans, dir, dentry, inode, 1, index);
5892

5893
	if (err) {
5894
		drop_inode = 1;
5895
	} else {
5896
		struct dentry *parent = dentry->d_parent;
5897
		err = btrfs_update_inode(trans, root, inode);
5898 5899
		if (err)
			goto fail;
5900
		d_instantiate(dentry, inode);
5901
		btrfs_log_new_name(trans, inode, NULL, parent);
5902
	}
C
Chris Mason 已提交
5903

5904
	btrfs_end_transaction(trans, root);
5905
fail:
C
Chris Mason 已提交
5906 5907 5908 5909
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
5910
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
5911 5912 5913
	return err;
}

5914
static int btrfs_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
C
Chris Mason 已提交
5915
{
5916
	struct inode *inode = NULL;
C
Chris Mason 已提交
5917 5918 5919 5920
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	int err = 0;
	int drop_on_err = 0;
5921
	u64 objectid = 0;
5922
	u64 index = 0;
C
Chris Mason 已提交
5923

J
Josef Bacik 已提交
5924 5925 5926 5927 5928
	/*
	 * 2 items for inode and ref
	 * 2 items for dir items
	 * 1 for xattr if selinux is on
	 */
5929 5930 5931
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
C
Chris Mason 已提交
5932

5933 5934 5935 5936
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_fail;

5937
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
5938
				dentry->d_name.len, btrfs_ino(dir), objectid,
5939
				S_IFDIR | mode, &index);
C
Chris Mason 已提交
5940 5941 5942 5943
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
		goto out_fail;
	}
5944

C
Chris Mason 已提交
5945
	drop_on_err = 1;
J
Josef Bacik 已提交
5946

5947
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
5948 5949 5950
	if (err)
		goto out_fail;

C
Chris Mason 已提交
5951 5952 5953
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

5954
	btrfs_i_size_write(inode, 0);
C
Chris Mason 已提交
5955 5956 5957
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		goto out_fail;
5958

5959 5960
	err = btrfs_add_link(trans, dir, inode, dentry->d_name.name,
			     dentry->d_name.len, 0, index);
C
Chris Mason 已提交
5961 5962
	if (err)
		goto out_fail;
5963

C
Chris Mason 已提交
5964 5965 5966 5967
	d_instantiate(dentry, inode);
	drop_on_err = 0;

out_fail:
5968
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
5969 5970
	if (drop_on_err)
		iput(inode);
5971
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
5972 5973 5974
	return err;
}

C
Chris Mason 已提交
5975 5976 5977 5978
/* helper for btfs_get_extent.  Given an existing extent in the tree,
 * and an extent that you want to insert, deal with overlap and insert
 * the new extent into the tree.
 */
5979 5980
static int merge_extent_mapping(struct extent_map_tree *em_tree,
				struct extent_map *existing,
5981 5982
				struct extent_map *em,
				u64 map_start, u64 map_len)
5983 5984 5985
{
	u64 start_diff;

5986 5987 5988 5989
	BUG_ON(map_start < em->start || map_start >= extent_map_end(em));
	start_diff = map_start - em->start;
	em->start = map_start;
	em->len = map_len;
C
Chris Mason 已提交
5990 5991
	if (em->block_start < EXTENT_MAP_LAST_BYTE &&
	    !test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
5992
		em->block_start += start_diff;
C
Chris Mason 已提交
5993 5994
		em->block_len -= start_diff;
	}
J
Josef Bacik 已提交
5995
	return add_extent_mapping(em_tree, em, 0);
5996 5997
}

C
Chris Mason 已提交
5998 5999 6000 6001 6002 6003 6004 6005 6006 6007 6008
static noinline int uncompress_inline(struct btrfs_path *path,
				      struct inode *inode, struct page *page,
				      size_t pg_offset, u64 extent_offset,
				      struct btrfs_file_extent_item *item)
{
	int ret;
	struct extent_buffer *leaf = path->nodes[0];
	char *tmp;
	size_t max_size;
	unsigned long inline_size;
	unsigned long ptr;
6009
	int compress_type;
C
Chris Mason 已提交
6010 6011

	WARN_ON(pg_offset != 0);
6012
	compress_type = btrfs_file_extent_compression(leaf, item);
C
Chris Mason 已提交
6013 6014 6015 6016
	max_size = btrfs_file_extent_ram_bytes(leaf, item);
	inline_size = btrfs_file_extent_inline_item_len(leaf,
					btrfs_item_nr(leaf, path->slots[0]));
	tmp = kmalloc(inline_size, GFP_NOFS);
6017 6018
	if (!tmp)
		return -ENOMEM;
C
Chris Mason 已提交
6019 6020 6021 6022
	ptr = btrfs_file_extent_inline_start(item);

	read_extent_buffer(leaf, tmp, ptr, inline_size);

6023
	max_size = min_t(unsigned long, PAGE_CACHE_SIZE, max_size);
6024 6025
	ret = btrfs_decompress(compress_type, tmp, page,
			       extent_offset, inline_size, max_size);
C
Chris Mason 已提交
6026
	if (ret) {
6027
		char *kaddr = kmap_atomic(page);
C
Chris Mason 已提交
6028 6029 6030 6031
		unsigned long copy_size = min_t(u64,
				  PAGE_CACHE_SIZE - pg_offset,
				  max_size - extent_offset);
		memset(kaddr + pg_offset, 0, copy_size);
6032
		kunmap_atomic(kaddr);
C
Chris Mason 已提交
6033 6034 6035 6036 6037
	}
	kfree(tmp);
	return 0;
}

C
Chris Mason 已提交
6038 6039
/*
 * a bit scary, this does extent mapping from logical file offset to the disk.
C
Chris Mason 已提交
6040 6041
 * the ugly parts come from merging extents from the disk with the in-ram
 * representation.  This gets more complex because of the data=ordered code,
C
Chris Mason 已提交
6042 6043 6044 6045
 * where the in-ram extents might be locked pending data=ordered completion.
 *
 * This also copies inline extents directly into the page.
 */
C
Chris Mason 已提交
6046

6047
struct extent_map *btrfs_get_extent(struct inode *inode, struct page *page,
6048
				    size_t pg_offset, u64 start, u64 len,
6049 6050 6051 6052
				    int create)
{
	int ret;
	int err = 0;
6053
	u64 bytenr;
6054 6055
	u64 extent_start = 0;
	u64 extent_end = 0;
L
Li Zefan 已提交
6056
	u64 objectid = btrfs_ino(inode);
6057
	u32 found_type;
6058
	struct btrfs_path *path = NULL;
6059 6060
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *item;
6061 6062
	struct extent_buffer *leaf;
	struct btrfs_key found_key;
6063 6064
	struct extent_map *em = NULL;
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
6065
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
6066
	struct btrfs_trans_handle *trans = NULL;
6067
	int compress_type;
6068 6069

again:
6070
	read_lock(&em_tree->lock);
6071
	em = lookup_extent_mapping(em_tree, start, len);
6072 6073
	if (em)
		em->bdev = root->fs_info->fs_devices->latest_bdev;
6074
	read_unlock(&em_tree->lock);
6075

6076
	if (em) {
6077 6078 6079
		if (em->start > start || em->start + em->len <= start)
			free_extent_map(em);
		else if (em->block_start == EXTENT_MAP_INLINE && page)
6080 6081 6082
			free_extent_map(em);
		else
			goto out;
6083
	}
6084
	em = alloc_extent_map();
6085
	if (!em) {
6086 6087
		err = -ENOMEM;
		goto out;
6088
	}
6089
	em->bdev = root->fs_info->fs_devices->latest_bdev;
6090
	em->start = EXTENT_MAP_HOLE;
6091
	em->orig_start = EXTENT_MAP_HOLE;
6092
	em->len = (u64)-1;
C
Chris Mason 已提交
6093
	em->block_len = (u64)-1;
6094 6095 6096

	if (!path) {
		path = btrfs_alloc_path();
J
Josef Bacik 已提交
6097 6098 6099 6100 6101 6102 6103 6104 6105
		if (!path) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * Chances are we'll be called again, so go ahead and do
		 * readahead
		 */
		path->reada = 1;
6106 6107
	}

6108 6109
	ret = btrfs_lookup_file_extent(trans, root, path,
				       objectid, start, trans != NULL);
6110 6111 6112 6113 6114 6115 6116 6117 6118 6119 6120
	if (ret < 0) {
		err = ret;
		goto out;
	}

	if (ret != 0) {
		if (path->slots[0] == 0)
			goto not_found;
		path->slots[0]--;
	}

6121 6122
	leaf = path->nodes[0];
	item = btrfs_item_ptr(leaf, path->slots[0],
6123 6124
			      struct btrfs_file_extent_item);
	/* are we inside the extent that was found? */
6125 6126 6127
	btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
	found_type = btrfs_key_type(&found_key);
	if (found_key.objectid != objectid ||
6128 6129 6130 6131
	    found_type != BTRFS_EXTENT_DATA_KEY) {
		goto not_found;
	}

6132 6133
	found_type = btrfs_file_extent_type(leaf, item);
	extent_start = found_key.offset;
6134
	compress_type = btrfs_file_extent_compression(leaf, item);
Y
Yan Zheng 已提交
6135 6136
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
6137
		extent_end = extent_start +
6138
		       btrfs_file_extent_num_bytes(leaf, item);
Y
Yan Zheng 已提交
6139 6140 6141
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
		size_t size;
		size = btrfs_file_extent_inline_len(leaf, item);
6142
		extent_end = ALIGN(extent_start + size, root->sectorsize);
Y
Yan Zheng 已提交
6143 6144 6145 6146 6147 6148 6149 6150 6151
	}

	if (start >= extent_end) {
		path->slots[0]++;
		if (path->slots[0] >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0) {
				err = ret;
				goto out;
6152
			}
Y
Yan Zheng 已提交
6153 6154 6155
			if (ret > 0)
				goto not_found;
			leaf = path->nodes[0];
6156
		}
Y
Yan Zheng 已提交
6157 6158 6159 6160 6161 6162 6163
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		if (found_key.objectid != objectid ||
		    found_key.type != BTRFS_EXTENT_DATA_KEY)
			goto not_found;
		if (start + len <= found_key.offset)
			goto not_found;
		em->start = start;
6164
		em->orig_start = start;
Y
Yan Zheng 已提交
6165 6166 6167 6168
		em->len = found_key.offset - start;
		goto not_found_em;
	}

J
Josef Bacik 已提交
6169
	em->ram_bytes = btrfs_file_extent_ram_bytes(leaf, item);
Y
Yan Zheng 已提交
6170 6171
	if (found_type == BTRFS_FILE_EXTENT_REG ||
	    found_type == BTRFS_FILE_EXTENT_PREALLOC) {
Y
Yan Zheng 已提交
6172 6173
		em->start = extent_start;
		em->len = extent_end - extent_start;
6174 6175
		em->orig_start = extent_start -
				 btrfs_file_extent_offset(leaf, item);
6176 6177
		em->orig_block_len = btrfs_file_extent_disk_num_bytes(leaf,
								      item);
6178 6179
		bytenr = btrfs_file_extent_disk_bytenr(leaf, item);
		if (bytenr == 0) {
6180
			em->block_start = EXTENT_MAP_HOLE;
6181 6182
			goto insert;
		}
6183
		if (compress_type != BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
6184
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
6185
			em->compress_type = compress_type;
C
Chris Mason 已提交
6186
			em->block_start = bytenr;
6187
			em->block_len = em->orig_block_len;
C
Chris Mason 已提交
6188 6189 6190 6191
		} else {
			bytenr += btrfs_file_extent_offset(leaf, item);
			em->block_start = bytenr;
			em->block_len = em->len;
Y
Yan Zheng 已提交
6192 6193
			if (found_type == BTRFS_FILE_EXTENT_PREALLOC)
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
C
Chris Mason 已提交
6194
		}
6195 6196
		goto insert;
	} else if (found_type == BTRFS_FILE_EXTENT_INLINE) {
6197
		unsigned long ptr;
6198
		char *map;
6199 6200 6201
		size_t size;
		size_t extent_offset;
		size_t copy_size;
6202

6203
		em->block_start = EXTENT_MAP_INLINE;
C
Chris Mason 已提交
6204
		if (!page || create) {
6205
			em->start = extent_start;
Y
Yan Zheng 已提交
6206
			em->len = extent_end - extent_start;
6207 6208
			goto out;
		}
6209

Y
Yan Zheng 已提交
6210 6211
		size = btrfs_file_extent_inline_len(leaf, item);
		extent_offset = page_offset(page) + pg_offset - extent_start;
6212
		copy_size = min_t(u64, PAGE_CACHE_SIZE - pg_offset,
6213 6214
				size - extent_offset);
		em->start = extent_start + extent_offset;
6215
		em->len = ALIGN(copy_size, root->sectorsize);
6216
		em->orig_block_len = em->len;
6217
		em->orig_start = em->start;
6218
		if (compress_type) {
C
Chris Mason 已提交
6219
			set_bit(EXTENT_FLAG_COMPRESSED, &em->flags);
6220 6221
			em->compress_type = compress_type;
		}
6222
		ptr = btrfs_file_extent_inline_start(item) + extent_offset;
6223
		if (create == 0 && !PageUptodate(page)) {
6224 6225
			if (btrfs_file_extent_compression(leaf, item) !=
			    BTRFS_COMPRESS_NONE) {
C
Chris Mason 已提交
6226 6227 6228
				ret = uncompress_inline(path, inode, page,
							pg_offset,
							extent_offset, item);
6229
				BUG_ON(ret); /* -ENOMEM */
C
Chris Mason 已提交
6230 6231 6232 6233
			} else {
				map = kmap(page);
				read_extent_buffer(leaf, map + pg_offset, ptr,
						   copy_size);
6234 6235 6236 6237 6238
				if (pg_offset + copy_size < PAGE_CACHE_SIZE) {
					memset(map + pg_offset + copy_size, 0,
					       PAGE_CACHE_SIZE - pg_offset -
					       copy_size);
				}
C
Chris Mason 已提交
6239 6240
				kunmap(page);
			}
6241 6242
			flush_dcache_page(page);
		} else if (create && PageUptodate(page)) {
6243
			BUG();
6244 6245 6246 6247
			if (!trans) {
				kunmap(page);
				free_extent_map(em);
				em = NULL;
C
Chris Mason 已提交
6248

6249
				btrfs_release_path(path);
6250
				trans = btrfs_join_transaction(root);
C
Chris Mason 已提交
6251

6252 6253
				if (IS_ERR(trans))
					return ERR_CAST(trans);
6254 6255
				goto again;
			}
C
Chris Mason 已提交
6256
			map = kmap(page);
6257
			write_extent_buffer(leaf, map + pg_offset, ptr,
6258
					    copy_size);
C
Chris Mason 已提交
6259
			kunmap(page);
6260
			btrfs_mark_buffer_dirty(leaf);
6261
		}
6262
		set_extent_uptodate(io_tree, em->start,
6263
				    extent_map_end(em) - 1, NULL, GFP_NOFS);
6264 6265
		goto insert;
	} else {
J
Julia Lawall 已提交
6266
		WARN(1, KERN_ERR "btrfs unknown found_type %d\n", found_type);
6267 6268 6269
	}
not_found:
	em->start = start;
6270
	em->orig_start = start;
6271
	em->len = len;
6272
not_found_em:
6273
	em->block_start = EXTENT_MAP_HOLE;
Y
Yan Zheng 已提交
6274
	set_bit(EXTENT_FLAG_VACANCY, &em->flags);
6275
insert:
6276
	btrfs_release_path(path);
6277
	if (em->start > start || extent_map_end(em) <= start) {
6278 6279 6280 6281 6282
		btrfs_err(root->fs_info, "bad extent! em: [%llu %llu] passed [%llu %llu]",
			(unsigned long long)em->start,
			(unsigned long long)em->len,
			(unsigned long long)start,
			(unsigned long long)len);
6283 6284 6285
		err = -EIO;
		goto out;
	}
6286 6287

	err = 0;
6288
	write_lock(&em_tree->lock);
J
Josef Bacik 已提交
6289
	ret = add_extent_mapping(em_tree, em, 0);
6290 6291 6292 6293
	/* it is possible that someone inserted the extent into the tree
	 * while we had the lock dropped.  It is also possible that
	 * an overlapping map exists in the tree
	 */
6294
	if (ret == -EEXIST) {
6295
		struct extent_map *existing;
6296 6297 6298

		ret = 0;

6299
		existing = lookup_extent_mapping(em_tree, start, len);
6300 6301 6302 6303 6304
		if (existing && (existing->start > start ||
		    existing->start + existing->len <= start)) {
			free_extent_map(existing);
			existing = NULL;
		}
6305 6306 6307 6308 6309
		if (!existing) {
			existing = lookup_extent_mapping(em_tree, em->start,
							 em->len);
			if (existing) {
				err = merge_extent_mapping(em_tree, existing,
6310 6311
							   em, start,
							   root->sectorsize);
6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324
				free_extent_map(existing);
				if (err) {
					free_extent_map(em);
					em = NULL;
				}
			} else {
				err = -EIO;
				free_extent_map(em);
				em = NULL;
			}
		} else {
			free_extent_map(em);
			em = existing;
6325
			err = 0;
6326 6327
		}
	}
6328
	write_unlock(&em_tree->lock);
6329
out:
6330

6331 6332
	if (em)
		trace_btrfs_get_extent(root, em);
6333

6334 6335
	if (path)
		btrfs_free_path(path);
6336 6337
	if (trans) {
		ret = btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
6338
		if (!err)
6339 6340 6341 6342 6343 6344
			err = ret;
	}
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
6345
	BUG_ON(!em); /* Error is always set */
6346 6347 6348
	return em;
}

6349 6350 6351 6352 6353 6354 6355 6356 6357 6358 6359 6360 6361 6362 6363 6364 6365
struct extent_map *btrfs_get_extent_fiemap(struct inode *inode, struct page *page,
					   size_t pg_offset, u64 start, u64 len,
					   int create)
{
	struct extent_map *em;
	struct extent_map *hole_em = NULL;
	u64 range_start = start;
	u64 end;
	u64 found;
	u64 found_end;
	int err = 0;

	em = btrfs_get_extent(inode, page, pg_offset, start, len, create);
	if (IS_ERR(em))
		return em;
	if (em) {
		/*
6366 6367 6368 6369
		 * if our em maps to
		 * -  a hole or
		 * -  a pre-alloc extent,
		 * there might actually be delalloc bytes behind it.
6370
		 */
6371 6372
		if (em->block_start != EXTENT_MAP_HOLE &&
		    !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
6373 6374 6375 6376 6377 6378 6379 6380 6381 6382 6383 6384 6385 6386 6387 6388 6389 6390 6391 6392 6393 6394 6395 6396 6397 6398 6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413
			return em;
		else
			hole_em = em;
	}

	/* check to see if we've wrapped (len == -1 or similar) */
	end = start + len;
	if (end < start)
		end = (u64)-1;
	else
		end -= 1;

	em = NULL;

	/* ok, we didn't find anything, lets look for delalloc */
	found = count_range_bits(&BTRFS_I(inode)->io_tree, &range_start,
				 end, len, EXTENT_DELALLOC, 1);
	found_end = range_start + found;
	if (found_end < range_start)
		found_end = (u64)-1;

	/*
	 * we didn't find anything useful, return
	 * the original results from get_extent()
	 */
	if (range_start > end || found_end <= start) {
		em = hole_em;
		hole_em = NULL;
		goto out;
	}

	/* adjust the range_start to make sure it doesn't
	 * go backwards from the start they passed in
	 */
	range_start = max(start,range_start);
	found = found_end - range_start;

	if (found > 0) {
		u64 hole_start = start;
		u64 hole_len = len;

6414
		em = alloc_extent_map();
6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432 6433 6434 6435 6436 6437 6438 6439 6440 6441 6442 6443 6444 6445 6446 6447 6448 6449 6450 6451 6452 6453
		if (!em) {
			err = -ENOMEM;
			goto out;
		}
		/*
		 * when btrfs_get_extent can't find anything it
		 * returns one huge hole
		 *
		 * make sure what it found really fits our range, and
		 * adjust to make sure it is based on the start from
		 * the caller
		 */
		if (hole_em) {
			u64 calc_end = extent_map_end(hole_em);

			if (calc_end <= start || (hole_em->start > end)) {
				free_extent_map(hole_em);
				hole_em = NULL;
			} else {
				hole_start = max(hole_em->start, start);
				hole_len = calc_end - hole_start;
			}
		}
		em->bdev = NULL;
		if (hole_em && range_start > hole_start) {
			/* our hole starts before our delalloc, so we
			 * have to return just the parts of the hole
			 * that go until  the delalloc starts
			 */
			em->len = min(hole_len,
				      range_start - hole_start);
			em->start = hole_start;
			em->orig_start = hole_start;
			/*
			 * don't adjust block start at all,
			 * it is fixed at EXTENT_MAP_HOLE
			 */
			em->block_start = hole_em->block_start;
			em->block_len = hole_len;
6454 6455
			if (test_bit(EXTENT_FLAG_PREALLOC, &hole_em->flags))
				set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
6456 6457 6458 6459 6460 6461 6462 6463 6464 6465 6466 6467 6468 6469 6470 6471 6472 6473 6474 6475
		} else {
			em->start = range_start;
			em->len = found;
			em->orig_start = range_start;
			em->block_start = EXTENT_MAP_DELALLOC;
			em->block_len = found;
		}
	} else if (hole_em) {
		return hole_em;
	}
out:

	free_extent_map(hole_em);
	if (err) {
		free_extent_map(em);
		return ERR_PTR(err);
	}
	return em;
}

6476 6477 6478 6479 6480
static struct extent_map *btrfs_new_extent_direct(struct inode *inode,
						  u64 start, u64 len)
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_trans_handle *trans;
6481
	struct extent_map *em;
6482 6483 6484 6485
	struct btrfs_key ins;
	u64 alloc_hint;
	int ret;

6486
	trans = btrfs_join_transaction(root);
6487 6488
	if (IS_ERR(trans))
		return ERR_CAST(trans);
6489 6490 6491 6492 6493

	trans->block_rsv = &root->fs_info->delalloc_block_rsv;

	alloc_hint = get_extent_allocation_hint(inode, start, len);
	ret = btrfs_reserve_extent(trans, root, len, root->sectorsize, 0,
6494
				   alloc_hint, &ins, 1);
6495 6496 6497 6498 6499
	if (ret) {
		em = ERR_PTR(ret);
		goto out;
	}

6500
	em = create_pinned_em(inode, start, ins.offset, start, ins.objectid,
J
Josef Bacik 已提交
6501
			      ins.offset, ins.offset, ins.offset, 0);
6502 6503
	if (IS_ERR(em))
		goto out;
6504 6505 6506 6507 6508 6509 6510 6511 6512 6513 6514 6515

	ret = btrfs_add_ordered_extent_dio(inode, start, ins.objectid,
					   ins.offset, ins.offset, 0);
	if (ret) {
		btrfs_free_reserved_extent(root, ins.objectid, ins.offset);
		em = ERR_PTR(ret);
	}
out:
	btrfs_end_transaction(trans, root);
	return em;
}

6516 6517 6518 6519 6520
/*
 * returns 1 when the nocow is safe, < 1 on error, 0 if the
 * block must be cow'd
 */
static noinline int can_nocow_odirect(struct btrfs_trans_handle *trans,
6521 6522 6523
				      struct inode *inode, u64 offset, u64 *len,
				      u64 *orig_start, u64 *orig_block_len,
				      u64 *ram_bytes)
6524 6525 6526 6527 6528 6529 6530 6531 6532 6533 6534 6535 6536 6537 6538 6539 6540 6541
{
	struct btrfs_path *path;
	int ret;
	struct extent_buffer *leaf;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_file_extent_item *fi;
	struct btrfs_key key;
	u64 disk_bytenr;
	u64 backref_offset;
	u64 extent_end;
	u64 num_bytes;
	int slot;
	int found_type;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

L
Li Zefan 已提交
6542
	ret = btrfs_lookup_file_extent(trans, root, path, btrfs_ino(inode),
6543 6544 6545 6546 6547 6548 6549 6550 6551 6552 6553 6554 6555 6556 6557 6558
				       offset, 0);
	if (ret < 0)
		goto out;

	slot = path->slots[0];
	if (ret == 1) {
		if (slot == 0) {
			/* can't find the item, must cow */
			ret = 0;
			goto out;
		}
		slot--;
	}
	ret = 0;
	leaf = path->nodes[0];
	btrfs_item_key_to_cpu(leaf, &key, slot);
L
Li Zefan 已提交
6559
	if (key.objectid != btrfs_ino(inode) ||
6560 6561 6562 6563 6564 6565 6566 6567 6568 6569 6570 6571 6572 6573 6574 6575 6576 6577 6578 6579
	    key.type != BTRFS_EXTENT_DATA_KEY) {
		/* not our file or wrong item type, must cow */
		goto out;
	}

	if (key.offset > offset) {
		/* Wrong offset, must cow */
		goto out;
	}

	fi = btrfs_item_ptr(leaf, slot, struct btrfs_file_extent_item);
	found_type = btrfs_file_extent_type(leaf, fi);
	if (found_type != BTRFS_FILE_EXTENT_REG &&
	    found_type != BTRFS_FILE_EXTENT_PREALLOC) {
		/* not a regular extent, must cow */
		goto out;
	}
	disk_bytenr = btrfs_file_extent_disk_bytenr(leaf, fi);
	backref_offset = btrfs_file_extent_offset(leaf, fi);

6580 6581 6582 6583
	*orig_start = key.offset - backref_offset;
	*orig_block_len = btrfs_file_extent_disk_num_bytes(leaf, fi);
	*ram_bytes = btrfs_file_extent_ram_bytes(leaf, fi);

6584
	extent_end = key.offset + btrfs_file_extent_num_bytes(leaf, fi);
6585
	if (extent_end < offset + *len) {
6586 6587 6588 6589 6590 6591 6592 6593 6594 6595 6596
		/* extent doesn't include our full range, must cow */
		goto out;
	}

	if (btrfs_extent_readonly(root, disk_bytenr))
		goto out;

	/*
	 * look for other files referencing this extent, if we
	 * find any we must cow
	 */
L
Li Zefan 已提交
6597
	if (btrfs_cross_ref_exist(trans, root, btrfs_ino(inode),
6598 6599 6600 6601 6602 6603 6604 6605 6606 6607 6608
				  key.offset - backref_offset, disk_bytenr))
		goto out;

	/*
	 * adjust disk_bytenr and num_bytes to cover just the bytes
	 * in this extent we are about to write.  If there
	 * are any csums in that range we have to cow in order
	 * to keep the csums correct
	 */
	disk_bytenr += backref_offset;
	disk_bytenr += offset - key.offset;
6609
	num_bytes = min(offset + *len, extent_end) - offset;
6610 6611 6612 6613 6614 6615
	if (csum_exist_in_range(root, disk_bytenr, num_bytes))
				goto out;
	/*
	 * all of the above have passed, it is safe to overwrite this extent
	 * without cow
	 */
6616
	*len = num_bytes;
6617 6618 6619 6620 6621 6622
	ret = 1;
out:
	btrfs_free_path(path);
	return ret;
}

6623 6624 6625 6626 6627 6628 6629 6630 6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641 6642 6643 6644 6645 6646 6647 6648 6649 6650 6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667 6668 6669 6670 6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683
static int lock_extent_direct(struct inode *inode, u64 lockstart, u64 lockend,
			      struct extent_state **cached_state, int writing)
{
	struct btrfs_ordered_extent *ordered;
	int ret = 0;

	while (1) {
		lock_extent_bits(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				 0, cached_state);
		/*
		 * We're concerned with the entire range that we're going to be
		 * doing DIO to, so we need to make sure theres no ordered
		 * extents in this range.
		 */
		ordered = btrfs_lookup_ordered_range(inode, lockstart,
						     lockend - lockstart + 1);

		/*
		 * We need to make sure there are no buffered pages in this
		 * range either, we could have raced between the invalidate in
		 * generic_file_direct_write and locking the extent.  The
		 * invalidate needs to happen so that reads after a write do not
		 * get stale data.
		 */
		if (!ordered && (!writing ||
		    !test_range_bit(&BTRFS_I(inode)->io_tree,
				    lockstart, lockend, EXTENT_UPTODATE, 0,
				    *cached_state)))
			break;

		unlock_extent_cached(&BTRFS_I(inode)->io_tree, lockstart, lockend,
				     cached_state, GFP_NOFS);

		if (ordered) {
			btrfs_start_ordered_extent(inode, ordered, 1);
			btrfs_put_ordered_extent(ordered);
		} else {
			/* Screw you mmap */
			ret = filemap_write_and_wait_range(inode->i_mapping,
							   lockstart,
							   lockend);
			if (ret)
				break;

			/*
			 * If we found a page that couldn't be invalidated just
			 * fall back to buffered.
			 */
			ret = invalidate_inode_pages2_range(inode->i_mapping,
					lockstart >> PAGE_CACHE_SHIFT,
					lockend >> PAGE_CACHE_SHIFT);
			if (ret)
				break;
		}

		cond_resched();
	}

	return ret;
}

6684 6685 6686
static struct extent_map *create_pinned_em(struct inode *inode, u64 start,
					   u64 len, u64 orig_start,
					   u64 block_start, u64 block_len,
J
Josef Bacik 已提交
6687 6688
					   u64 orig_block_len, u64 ram_bytes,
					   int type)
6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700 6701
{
	struct extent_map_tree *em_tree;
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

	em_tree = &BTRFS_I(inode)->extent_tree;
	em = alloc_extent_map();
	if (!em)
		return ERR_PTR(-ENOMEM);

	em->start = start;
	em->orig_start = orig_start;
6702 6703
	em->mod_start = start;
	em->mod_len = len;
6704 6705 6706 6707
	em->len = len;
	em->block_len = block_len;
	em->block_start = block_start;
	em->bdev = root->fs_info->fs_devices->latest_bdev;
6708
	em->orig_block_len = orig_block_len;
J
Josef Bacik 已提交
6709
	em->ram_bytes = ram_bytes;
6710
	em->generation = -1;
6711 6712
	set_bit(EXTENT_FLAG_PINNED, &em->flags);
	if (type == BTRFS_ORDERED_PREALLOC)
6713
		set_bit(EXTENT_FLAG_FILLING, &em->flags);
6714 6715 6716 6717 6718

	do {
		btrfs_drop_extent_cache(inode, em->start,
				em->start + em->len - 1, 0);
		write_lock(&em_tree->lock);
J
Josef Bacik 已提交
6719
		ret = add_extent_mapping(em_tree, em, 1);
6720 6721 6722 6723 6724 6725 6726 6727 6728 6729 6730 6731
		write_unlock(&em_tree->lock);
	} while (ret == -EEXIST);

	if (ret) {
		free_extent_map(em);
		return ERR_PTR(ret);
	}

	return em;
}


6732 6733 6734 6735 6736
static int btrfs_get_blocks_direct(struct inode *inode, sector_t iblock,
				   struct buffer_head *bh_result, int create)
{
	struct extent_map *em;
	struct btrfs_root *root = BTRFS_I(inode)->root;
6737
	struct extent_state *cached_state = NULL;
6738
	u64 start = iblock << inode->i_blkbits;
6739
	u64 lockstart, lockend;
6740
	u64 len = bh_result->b_size;
6741
	struct btrfs_trans_handle *trans;
6742
	int unlock_bits = EXTENT_LOCKED;
6743
	int ret = 0;
6744

6745
	if (create)
6746
		unlock_bits |= EXTENT_DELALLOC | EXTENT_DIRTY;
6747
	else
6748
		len = min_t(u64, len, root->sectorsize);
6749

6750 6751 6752
	lockstart = start;
	lockend = start + len - 1;

6753 6754 6755 6756 6757 6758 6759
	/*
	 * If this errors out it's because we couldn't invalidate pagecache for
	 * this range and we need to fallback to buffered.
	 */
	if (lock_extent_direct(inode, lockstart, lockend, &cached_state, create))
		return -ENOTBLK;

6760
	em = btrfs_get_extent(inode, NULL, 0, start, len, 0);
6761 6762 6763 6764
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto unlock_err;
	}
6765 6766 6767 6768 6769 6770 6771 6772 6773 6774 6775 6776 6777 6778 6779 6780 6781 6782

	/*
	 * Ok for INLINE and COMPRESSED extents we need to fallback on buffered
	 * io.  INLINE is special, and we could probably kludge it in here, but
	 * it's still buffered so for safety lets just fall back to the generic
	 * buffered path.
	 *
	 * For COMPRESSED we _have_ to read the entire extent in so we can
	 * decompress it, so there will be buffering required no matter what we
	 * do, so go ahead and fallback to buffered.
	 *
	 * We return -ENOTBLK because thats what makes DIO go ahead and go back
	 * to buffered IO.  Don't blame me, this is the price we pay for using
	 * the generic code.
	 */
	if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags) ||
	    em->block_start == EXTENT_MAP_INLINE) {
		free_extent_map(em);
6783 6784
		ret = -ENOTBLK;
		goto unlock_err;
6785 6786 6787 6788 6789 6790
	}

	/* Just a good old fashioned hole, return */
	if (!create && (em->block_start == EXTENT_MAP_HOLE ||
			test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
		free_extent_map(em);
6791
		goto unlock_err;
6792 6793 6794 6795 6796 6797 6798 6799 6800 6801 6802
	}

	/*
	 * We don't allocate a new extent in the following cases
	 *
	 * 1) The inode is marked as NODATACOW.  In this case we'll just use the
	 * existing extent.
	 * 2) The extent is marked as PREALLOC.  We're good to go here and can
	 * just use the extent.
	 *
	 */
6803
	if (!create) {
6804 6805 6806
		len = min(len, em->len - (start - em->start));
		lockstart = start + len;
		goto unlock;
6807
	}
6808 6809 6810 6811 6812 6813

	if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags) ||
	    ((BTRFS_I(inode)->flags & BTRFS_INODE_NODATACOW) &&
	     em->block_start != EXTENT_MAP_HOLE)) {
		int type;
		int ret;
6814
		u64 block_start, orig_start, orig_block_len, ram_bytes;
6815 6816 6817 6818 6819

		if (test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			type = BTRFS_ORDERED_PREALLOC;
		else
			type = BTRFS_ORDERED_NOCOW;
6820
		len = min(len, em->len - (start - em->start));
6821
		block_start = em->block_start + (start - em->start);
6822 6823 6824 6825 6826 6827

		/*
		 * we're not going to log anything, but we do need
		 * to make sure the current transaction stays open
		 * while we look for nocow cross refs
		 */
6828
		trans = btrfs_join_transaction(root);
6829
		if (IS_ERR(trans))
6830 6831
			goto must_cow;

6832 6833
		if (can_nocow_odirect(trans, inode, start, &len, &orig_start,
				      &orig_block_len, &ram_bytes) == 1) {
6834 6835 6836 6837
			if (type == BTRFS_ORDERED_PREALLOC) {
				free_extent_map(em);
				em = create_pinned_em(inode, start, len,
						       orig_start,
6838
						       block_start, len,
J
Josef Bacik 已提交
6839 6840
						       orig_block_len,
						       ram_bytes, type);
6841 6842 6843 6844 6845 6846
				if (IS_ERR(em)) {
					btrfs_end_transaction(trans, root);
					goto unlock_err;
				}
			}

6847 6848 6849 6850 6851
			ret = btrfs_add_ordered_extent_dio(inode, start,
					   block_start, len, len, type);
			btrfs_end_transaction(trans, root);
			if (ret) {
				free_extent_map(em);
6852
				goto unlock_err;
6853 6854
			}
			goto unlock;
6855
		}
6856
		btrfs_end_transaction(trans, root);
6857
	}
6858 6859 6860 6861 6862 6863
must_cow:
	/*
	 * this will cow the extent, reset the len in case we changed
	 * it above
	 */
	len = bh_result->b_size;
6864 6865
	free_extent_map(em);
	em = btrfs_new_extent_direct(inode, start, len);
6866 6867 6868 6869
	if (IS_ERR(em)) {
		ret = PTR_ERR(em);
		goto unlock_err;
	}
6870 6871
	len = min(len, em->len - (start - em->start));
unlock:
6872 6873
	bh_result->b_blocknr = (em->block_start + (start - em->start)) >>
		inode->i_blkbits;
6874
	bh_result->b_size = len;
6875 6876
	bh_result->b_bdev = em->bdev;
	set_buffer_mapped(bh_result);
6877 6878 6879 6880 6881 6882 6883 6884 6885 6886
	if (create) {
		if (!test_bit(EXTENT_FLAG_PREALLOC, &em->flags))
			set_buffer_new(bh_result);

		/*
		 * Need to update the i_size under the extent lock so buffered
		 * readers will get the updated i_size when we unlock.
		 */
		if (start + len > i_size_read(inode))
			i_size_write(inode, start + len);
6887

6888 6889 6890 6891
		spin_lock(&BTRFS_I(inode)->lock);
		BTRFS_I(inode)->outstanding_extents++;
		spin_unlock(&BTRFS_I(inode)->lock);

6892 6893 6894 6895
		ret = set_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
				     lockstart + len - 1, EXTENT_DELALLOC, NULL,
				     &cached_state, GFP_NOFS);
		BUG_ON(ret);
6896
	}
6897

6898 6899 6900 6901 6902
	/*
	 * In the case of write we need to clear and unlock the entire range,
	 * in the case of read we need to unlock only the end area that we
	 * aren't using if there is any left over space.
	 */
L
Liu Bo 已提交
6903
	if (lockstart < lockend) {
6904 6905 6906
		clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart,
				 lockend, unlock_bits, 1, 0,
				 &cached_state, GFP_NOFS);
L
Liu Bo 已提交
6907
	} else {
6908
		free_extent_state(cached_state);
L
Liu Bo 已提交
6909
	}
6910

6911 6912 6913
	free_extent_map(em);

	return 0;
6914 6915 6916 6917 6918

unlock_err:
	clear_extent_bit(&BTRFS_I(inode)->io_tree, lockstart, lockend,
			 unlock_bits, 1, 0, &cached_state, GFP_NOFS);
	return ret;
6919 6920 6921 6922 6923 6924 6925 6926
}

struct btrfs_dio_private {
	struct inode *inode;
	u64 logical_offset;
	u64 disk_bytenr;
	u64 bytes;
	void *private;
M
Miao Xie 已提交
6927 6928 6929 6930 6931 6932 6933

	/* number of bios pending for this dio */
	atomic_t pending_bios;

	/* IO errors */
	int errors;

6934
	/* orig_bio is our btrfs_io_bio */
M
Miao Xie 已提交
6935
	struct bio *orig_bio;
6936 6937 6938

	/* dio_bio came from fs/direct-io.c */
	struct bio *dio_bio;
6939 6940 6941 6942
};

static void btrfs_endio_direct_read(struct bio *bio, int err)
{
M
Miao Xie 已提交
6943
	struct btrfs_dio_private *dip = bio->bi_private;
6944 6945 6946
	struct bio_vec *bvec_end = bio->bi_io_vec + bio->bi_vcnt - 1;
	struct bio_vec *bvec = bio->bi_io_vec;
	struct inode *inode = dip->inode;
6947
	struct btrfs_root *root = BTRFS_I(inode)->root;
6948
	struct bio *dio_bio;
6949 6950 6951 6952 6953 6954 6955 6956
	u64 start;

	start = dip->logical_offset;
	do {
		if (!(BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM)) {
			struct page *page = bvec->bv_page;
			char *kaddr;
			u32 csum = ~(u32)0;
6957
			u64 private = ~(u32)0;
6958 6959
			unsigned long flags;

6960 6961 6962
			if (get_state_private(&BTRFS_I(inode)->io_tree,
					      start, &private))
				goto failed;
6963
			local_irq_save(flags);
6964
			kaddr = kmap_atomic(page);
6965
			csum = btrfs_csum_data(kaddr + bvec->bv_offset,
6966 6967
					       csum, bvec->bv_len);
			btrfs_csum_final(csum, (char *)&csum);
6968
			kunmap_atomic(kaddr);
6969 6970 6971
			local_irq_restore(flags);

			flush_dcache_page(bvec->bv_page);
6972 6973
			if (csum != private) {
failed:
6974 6975 6976 6977
				btrfs_err(root->fs_info, "csum failed ino %llu off %llu csum %u private %u",
					(unsigned long long)btrfs_ino(inode),
					(unsigned long long)start,
					csum, (unsigned)private);
6978 6979 6980 6981 6982 6983 6984 6985 6986
				err = -EIO;
			}
		}

		start += bvec->bv_len;
		bvec++;
	} while (bvec <= bvec_end);

	unlock_extent(&BTRFS_I(inode)->io_tree, dip->logical_offset,
6987
		      dip->logical_offset + dip->bytes - 1);
6988
	dio_bio = dip->dio_bio;
6989 6990

	kfree(dip);
6991 6992 6993

	/* If we had a csum failure make sure to clear the uptodate flag */
	if (err)
6994 6995 6996
		clear_bit(BIO_UPTODATE, &dio_bio->bi_flags);
	dio_end_io(dio_bio, err);
	bio_put(bio);
6997 6998 6999 7000 7001 7002 7003 7004
}

static void btrfs_endio_direct_write(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_ordered_extent *ordered = NULL;
7005 7006
	u64 ordered_offset = dip->logical_offset;
	u64 ordered_bytes = dip->bytes;
7007
	struct bio *dio_bio;
7008 7009 7010 7011
	int ret;

	if (err)
		goto out_done;
7012 7013 7014
again:
	ret = btrfs_dec_test_first_ordered_pending(inode, &ordered,
						   &ordered_offset,
7015
						   ordered_bytes, !err);
7016
	if (!ret)
7017
		goto out_test;
7018

7019 7020 7021 7022
	ordered->work.func = finish_ordered_fn;
	ordered->work.flags = 0;
	btrfs_queue_worker(&root->fs_info->endio_write_workers,
			   &ordered->work);
7023 7024 7025 7026 7027 7028 7029 7030
out_test:
	/*
	 * our bio might span multiple ordered extents.  If we haven't
	 * completed the accounting for the whole dio, go back and try again
	 */
	if (ordered_offset < dip->logical_offset + dip->bytes) {
		ordered_bytes = dip->logical_offset + dip->bytes -
			ordered_offset;
7031
		ordered = NULL;
7032 7033
		goto again;
	}
7034
out_done:
7035
	dio_bio = dip->dio_bio;
7036 7037

	kfree(dip);
7038 7039 7040

	/* If we had an error make sure to clear the uptodate flag */
	if (err)
7041 7042 7043
		clear_bit(BIO_UPTODATE, &dio_bio->bi_flags);
	dio_end_io(dio_bio, err);
	bio_put(bio);
7044 7045
}

7046 7047 7048 7049 7050 7051 7052
static int __btrfs_submit_bio_start_direct_io(struct inode *inode, int rw,
				    struct bio *bio, int mirror_num,
				    unsigned long bio_flags, u64 offset)
{
	int ret;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	ret = btrfs_csum_one_bio(root, inode, bio, offset, 1);
7053
	BUG_ON(ret); /* -ENOMEM */
7054 7055 7056
	return 0;
}

M
Miao Xie 已提交
7057 7058 7059 7060 7061
static void btrfs_end_dio_bio(struct bio *bio, int err)
{
	struct btrfs_dio_private *dip = bio->bi_private;

	if (err) {
L
Li Zefan 已提交
7062
		printk(KERN_ERR "btrfs direct IO failed ino %llu rw %lu "
J
Jan Beulich 已提交
7063
		      "sector %#Lx len %u err no %d\n",
L
Li Zefan 已提交
7064
		      (unsigned long long)btrfs_ino(dip->inode), bio->bi_rw,
J
Jan Beulich 已提交
7065
		      (unsigned long long)bio->bi_sector, bio->bi_size, err);
M
Miao Xie 已提交
7066 7067 7068 7069 7070 7071 7072 7073 7074 7075 7076 7077 7078
		dip->errors = 1;

		/*
		 * before atomic variable goto zero, we must make sure
		 * dip->errors is perceived to be set.
		 */
		smp_mb__before_atomic_dec();
	}

	/* if there are more bios still pending for this dio, just exit */
	if (!atomic_dec_and_test(&dip->pending_bios))
		goto out;

7079
	if (dip->errors) {
M
Miao Xie 已提交
7080
		bio_io_error(dip->orig_bio);
7081 7082
	} else {
		set_bit(BIO_UPTODATE, &dip->dio_bio->bi_flags);
M
Miao Xie 已提交
7083 7084 7085 7086 7087 7088 7089 7090 7091 7092 7093 7094 7095 7096 7097
		bio_endio(dip->orig_bio, 0);
	}
out:
	bio_put(bio);
}

static struct bio *btrfs_dio_bio_alloc(struct block_device *bdev,
				       u64 first_sector, gfp_t gfp_flags)
{
	int nr_vecs = bio_get_nr_vecs(bdev);
	return btrfs_bio_alloc(bdev, first_sector, nr_vecs, gfp_flags);
}

static inline int __btrfs_submit_dio_bio(struct bio *bio, struct inode *inode,
					 int rw, u64 file_offset, int skip_sum,
7098
					 int async_submit)
M
Miao Xie 已提交
7099 7100 7101 7102 7103
{
	int write = rw & REQ_WRITE;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	int ret;

7104 7105 7106
	if (async_submit)
		async_submit = !atomic_read(&BTRFS_I(inode)->sync_writers);

M
Miao Xie 已提交
7107
	bio_get(bio);
7108 7109 7110 7111 7112 7113

	if (!write) {
		ret = btrfs_bio_wq_end_io(root->fs_info, bio, 0);
		if (ret)
			goto err;
	}
M
Miao Xie 已提交
7114

7115 7116 7117 7118
	if (skip_sum)
		goto map;

	if (write && async_submit) {
M
Miao Xie 已提交
7119 7120 7121 7122 7123 7124
		ret = btrfs_wq_submit_bio(root->fs_info,
				   inode, rw, bio, 0, 0,
				   file_offset,
				   __btrfs_submit_bio_start_direct_io,
				   __btrfs_submit_bio_done);
		goto err;
7125 7126 7127 7128 7129 7130 7131 7132
	} else if (write) {
		/*
		 * If we aren't doing async submit, calculate the csum of the
		 * bio now.
		 */
		ret = btrfs_csum_one_bio(root, inode, bio, file_offset, 1);
		if (ret)
			goto err;
7133
	} else if (!skip_sum) {
7134
		ret = btrfs_lookup_bio_sums_dio(root, inode, bio, file_offset);
7135 7136 7137
		if (ret)
			goto err;
	}
M
Miao Xie 已提交
7138

7139 7140
map:
	ret = btrfs_map_bio(root, rw, bio, 0, async_submit);
M
Miao Xie 已提交
7141 7142 7143 7144 7145 7146 7147 7148 7149 7150 7151 7152 7153 7154 7155 7156 7157 7158 7159
err:
	bio_put(bio);
	return ret;
}

static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
				    int skip_sum)
{
	struct inode *inode = dip->inode;
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct bio *bio;
	struct bio *orig_bio = dip->orig_bio;
	struct bio_vec *bvec = orig_bio->bi_io_vec;
	u64 start_sector = orig_bio->bi_sector;
	u64 file_offset = dip->logical_offset;
	u64 submit_len = 0;
	u64 map_length;
	int nr_pages = 0;
	int ret = 0;
7160
	int async_submit = 0;
M
Miao Xie 已提交
7161 7162

	map_length = orig_bio->bi_size;
D
David Woodhouse 已提交
7163
	ret = btrfs_map_block(root->fs_info, rw, start_sector << 9,
M
Miao Xie 已提交
7164 7165
			      &map_length, NULL, 0);
	if (ret) {
7166
		bio_put(orig_bio);
M
Miao Xie 已提交
7167 7168
		return -EIO;
	}
7169 7170 7171 7172 7173
	if (map_length >= orig_bio->bi_size) {
		bio = orig_bio;
		goto submit;
	}

D
David Woodhouse 已提交
7174 7175 7176 7177 7178 7179 7180
	/* async crcs make it difficult to collect full stripe writes. */
	if (btrfs_get_alloc_profile(root, 1) &
	    (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6))
		async_submit = 0;
	else
		async_submit = 1;

7181 7182 7183 7184 7185 7186 7187
	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
	if (!bio)
		return -ENOMEM;
	bio->bi_private = dip;
	bio->bi_end_io = btrfs_end_dio_bio;
	atomic_inc(&dip->pending_bios);

M
Miao Xie 已提交
7188 7189 7190 7191 7192 7193 7194 7195 7196 7197 7198 7199 7200
	while (bvec <= (orig_bio->bi_io_vec + orig_bio->bi_vcnt - 1)) {
		if (unlikely(map_length < submit_len + bvec->bv_len ||
		    bio_add_page(bio, bvec->bv_page, bvec->bv_len,
				 bvec->bv_offset) < bvec->bv_len)) {
			/*
			 * inc the count before we submit the bio so
			 * we know the end IO handler won't happen before
			 * we inc the count. Otherwise, the dip might get freed
			 * before we're done setting it up
			 */
			atomic_inc(&dip->pending_bios);
			ret = __btrfs_submit_dio_bio(bio, inode, rw,
						     file_offset, skip_sum,
7201
						     async_submit);
M
Miao Xie 已提交
7202 7203 7204 7205 7206 7207 7208 7209 7210 7211 7212 7213 7214 7215 7216 7217 7218 7219 7220 7221
			if (ret) {
				bio_put(bio);
				atomic_dec(&dip->pending_bios);
				goto out_err;
			}

			start_sector += submit_len >> 9;
			file_offset += submit_len;

			submit_len = 0;
			nr_pages = 0;

			bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev,
						  start_sector, GFP_NOFS);
			if (!bio)
				goto out_err;
			bio->bi_private = dip;
			bio->bi_end_io = btrfs_end_dio_bio;

			map_length = orig_bio->bi_size;
D
David Woodhouse 已提交
7222
			ret = btrfs_map_block(root->fs_info, rw,
7223
					      start_sector << 9,
M
Miao Xie 已提交
7224 7225 7226 7227 7228 7229 7230 7231 7232 7233 7234 7235
					      &map_length, NULL, 0);
			if (ret) {
				bio_put(bio);
				goto out_err;
			}
		} else {
			submit_len += bvec->bv_len;
			nr_pages ++;
			bvec++;
		}
	}

7236
submit:
M
Miao Xie 已提交
7237
	ret = __btrfs_submit_dio_bio(bio, inode, rw, file_offset, skip_sum,
7238
				     async_submit);
M
Miao Xie 已提交
7239 7240 7241 7242 7243 7244 7245 7246 7247 7248 7249 7250 7251 7252 7253 7254 7255 7256
	if (!ret)
		return 0;

	bio_put(bio);
out_err:
	dip->errors = 1;
	/*
	 * before atomic variable goto zero, we must
	 * make sure dip->errors is perceived to be set.
	 */
	smp_mb__before_atomic_dec();
	if (atomic_dec_and_test(&dip->pending_bios))
		bio_io_error(dip->orig_bio);

	/* bio_end_io() will handle error, so we needn't return it */
	return 0;
}

7257 7258
static void btrfs_submit_direct(int rw, struct bio *dio_bio,
				struct inode *inode, loff_t file_offset)
7259 7260 7261
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_dio_private *dip;
7262 7263
	struct bio_vec *bvec = dio_bio->bi_io_vec;
	struct bio *io_bio;
7264
	int skip_sum;
7265
	int write = rw & REQ_WRITE;
7266 7267 7268 7269
	int ret = 0;

	skip_sum = BTRFS_I(inode)->flags & BTRFS_INODE_NODATASUM;

7270 7271 7272 7273 7274 7275 7276
	io_bio = btrfs_bio_clone(dio_bio, GFP_NOFS);

	if (!io_bio) {
		ret = -ENOMEM;
		goto free_ordered;
	}

7277 7278 7279
	dip = kmalloc(sizeof(*dip), GFP_NOFS);
	if (!dip) {
		ret = -ENOMEM;
7280
		goto free_io_bio;
7281 7282
	}

7283 7284
	dip->private = dio_bio->bi_private;
	io_bio->bi_private = dio_bio->bi_private;
7285 7286 7287 7288 7289 7290 7291
	dip->inode = inode;
	dip->logical_offset = file_offset;

	dip->bytes = 0;
	do {
		dip->bytes += bvec->bv_len;
		bvec++;
7292
	} while (bvec <= (dio_bio->bi_io_vec + dio_bio->bi_vcnt - 1));
7293

7294 7295
	dip->disk_bytenr = (u64)dio_bio->bi_sector << 9;
	io_bio->bi_private = dip;
M
Miao Xie 已提交
7296
	dip->errors = 0;
7297 7298
	dip->orig_bio = io_bio;
	dip->dio_bio = dio_bio;
M
Miao Xie 已提交
7299
	atomic_set(&dip->pending_bios, 0);
7300 7301

	if (write)
7302
		io_bio->bi_end_io = btrfs_endio_direct_write;
7303
	else
7304
		io_bio->bi_end_io = btrfs_endio_direct_read;
7305

M
Miao Xie 已提交
7306 7307
	ret = btrfs_submit_direct_hook(rw, dip, skip_sum);
	if (!ret)
7308
		return;
7309 7310 7311 7312

free_io_bio:
	bio_put(io_bio);

7313 7314 7315 7316 7317 7318 7319
free_ordered:
	/*
	 * If this is a write, we need to clean up the reserved space and kill
	 * the ordered extent.
	 */
	if (write) {
		struct btrfs_ordered_extent *ordered;
7320
		ordered = btrfs_lookup_ordered_extent(inode, file_offset);
7321 7322 7323 7324 7325 7326 7327
		if (!test_bit(BTRFS_ORDERED_PREALLOC, &ordered->flags) &&
		    !test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags))
			btrfs_free_reserved_extent(root, ordered->start,
						   ordered->disk_len);
		btrfs_put_ordered_extent(ordered);
		btrfs_put_ordered_extent(ordered);
	}
7328
	bio_endio(dio_bio, ret);
7329 7330
}

C
Chris Mason 已提交
7331 7332 7333 7334 7335
static ssize_t check_direct_IO(struct btrfs_root *root, int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
	int seg;
7336
	int i;
C
Chris Mason 已提交
7337 7338 7339 7340 7341 7342 7343 7344 7345 7346 7347 7348 7349 7350
	size_t size;
	unsigned long addr;
	unsigned blocksize_mask = root->sectorsize - 1;
	ssize_t retval = -EINVAL;
	loff_t end = offset;

	if (offset & blocksize_mask)
		goto out;

	/* Check the memory alignment.  Blocks cannot straddle pages */
	for (seg = 0; seg < nr_segs; seg++) {
		addr = (unsigned long)iov[seg].iov_base;
		size = iov[seg].iov_len;
		end += size;
7351
		if ((addr & blocksize_mask) || (size & blocksize_mask))
C
Chris Mason 已提交
7352
			goto out;
7353 7354 7355 7356 7357 7358 7359 7360 7361 7362 7363 7364 7365 7366

		/* If this is a write we don't need to check anymore */
		if (rw & WRITE)
			continue;

		/*
		 * Check to make sure we don't have duplicate iov_base's in this
		 * iovec, if so return EINVAL, otherwise we'll get csum errors
		 * when reading back.
		 */
		for (i = seg + 1; i < nr_segs; i++) {
			if (iov[seg].iov_base == iov[i].iov_base)
				goto out;
		}
C
Chris Mason 已提交
7367 7368 7369 7370 7371
	}
	retval = 0;
out:
	return retval;
}
7372

7373 7374 7375 7376
static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
			const struct iovec *iov, loff_t offset,
			unsigned long nr_segs)
{
7377 7378
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_mapping->host;
7379
	size_t count = 0;
7380
	int flags = 0;
M
Miao Xie 已提交
7381 7382
	bool wakeup = true;
	bool relock = false;
7383
	ssize_t ret;
7384

C
Chris Mason 已提交
7385
	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov,
7386
			    offset, nr_segs))
C
Chris Mason 已提交
7387
		return 0;
7388

M
Miao Xie 已提交
7389 7390 7391
	atomic_inc(&inode->i_dio_count);
	smp_mb__after_atomic_inc();

7392 7393
	if (rw & WRITE) {
		count = iov_length(iov, nr_segs);
M
Miao Xie 已提交
7394 7395 7396 7397 7398 7399 7400 7401 7402
		/*
		 * If the write DIO is beyond the EOF, we need update
		 * the isize, but it is protected by i_mutex. So we can
		 * not unlock the i_mutex at this case.
		 */
		if (offset + count <= inode->i_size) {
			mutex_unlock(&inode->i_mutex);
			relock = true;
		}
7403 7404
		ret = btrfs_delalloc_reserve_space(inode, count);
		if (ret)
M
Miao Xie 已提交
7405 7406 7407 7408 7409 7410
			goto out;
	} else if (unlikely(test_bit(BTRFS_INODE_READDIO_NEED_LOCK,
				     &BTRFS_I(inode)->runtime_flags))) {
		inode_dio_done(inode);
		flags = DIO_LOCKING | DIO_SKIP_HOLES;
		wakeup = false;
7411 7412 7413 7414 7415
	}

	ret = __blockdev_direct_IO(rw, iocb, inode,
			BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev,
			iov, offset, nr_segs, btrfs_get_blocks_direct, NULL,
7416
			btrfs_submit_direct, flags);
7417 7418 7419
	if (rw & WRITE) {
		if (ret < 0 && ret != -EIOCBQUEUED)
			btrfs_delalloc_release_space(inode, count);
7420
		else if (ret >= 0 && (size_t)ret < count)
7421 7422
			btrfs_delalloc_release_space(inode,
						     count - (size_t)ret);
7423 7424
		else
			btrfs_delalloc_release_metadata(inode, 0);
7425
	}
M
Miao Xie 已提交
7426
out:
7427 7428
	if (wakeup)
		inode_dio_done(inode);
M
Miao Xie 已提交
7429 7430
	if (relock)
		mutex_lock(&inode->i_mutex);
7431 7432

	return ret;
7433 7434
}

T
Tsutomu Itoh 已提交
7435 7436
#define BTRFS_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC)

Y
Yehuda Sadeh 已提交
7437 7438 7439
static int btrfs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
		__u64 start, __u64 len)
{
T
Tsutomu Itoh 已提交
7440 7441 7442 7443 7444 7445
	int	ret;

	ret = fiemap_check_flags(fieinfo, BTRFS_FIEMAP_FLAGS);
	if (ret)
		return ret;

7446
	return extent_fiemap(inode, fieinfo, start, len, btrfs_get_extent_fiemap);
Y
Yehuda Sadeh 已提交
7447 7448
}

7449
int btrfs_readpage(struct file *file, struct page *page)
C
Chris Mason 已提交
7450
{
7451 7452
	struct extent_io_tree *tree;
	tree = &BTRFS_I(page->mapping->host)->io_tree;
7453
	return extent_read_full_page(tree, page, btrfs_get_extent, 0);
C
Chris Mason 已提交
7454
}
7455

7456
static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
C
Chris Mason 已提交
7457
{
7458
	struct extent_io_tree *tree;
7459 7460 7461 7462 7463 7464 7465


	if (current->flags & PF_MEMALLOC) {
		redirty_page_for_writepage(wbc, page);
		unlock_page(page);
		return 0;
	}
7466
	tree = &BTRFS_I(page->mapping->host)->io_tree;
7467
	return extent_write_full_page(tree, page, btrfs_get_extent, wbc);
C
Chris Mason 已提交
7468 7469
}

7470 7471
static int btrfs_writepages(struct address_space *mapping,
			    struct writeback_control *wbc)
C
Chris Mason 已提交
7472
{
7473
	struct extent_io_tree *tree;
7474

7475
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
7476 7477 7478
	return extent_writepages(tree, mapping, btrfs_get_extent, wbc);
}

C
Chris Mason 已提交
7479 7480 7481 7482
static int
btrfs_readpages(struct file *file, struct address_space *mapping,
		struct list_head *pages, unsigned nr_pages)
{
7483 7484
	struct extent_io_tree *tree;
	tree = &BTRFS_I(mapping->host)->io_tree;
C
Chris Mason 已提交
7485 7486 7487
	return extent_readpages(tree, mapping, pages, nr_pages,
				btrfs_get_extent);
}
7488
static int __btrfs_releasepage(struct page *page, gfp_t gfp_flags)
C
Chris Mason 已提交
7489
{
7490 7491
	struct extent_io_tree *tree;
	struct extent_map_tree *map;
7492
	int ret;
7493

7494 7495
	tree = &BTRFS_I(page->mapping->host)->io_tree;
	map = &BTRFS_I(page->mapping->host)->extent_tree;
7496
	ret = try_release_extent_mapping(map, tree, page, gfp_flags);
7497 7498 7499 7500
	if (ret == 1) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
C
Chris Mason 已提交
7501
	}
7502
	return ret;
C
Chris Mason 已提交
7503 7504
}

7505 7506
static int btrfs_releasepage(struct page *page, gfp_t gfp_flags)
{
7507 7508
	if (PageWriteback(page) || PageDirty(page))
		return 0;
7509
	return __btrfs_releasepage(page, gfp_flags & GFP_NOFS);
7510 7511
}

7512
static void btrfs_invalidatepage(struct page *page, unsigned long offset)
C
Chris Mason 已提交
7513
{
7514
	struct inode *inode = page->mapping->host;
7515
	struct extent_io_tree *tree;
7516
	struct btrfs_ordered_extent *ordered;
7517
	struct extent_state *cached_state = NULL;
7518 7519
	u64 page_start = page_offset(page);
	u64 page_end = page_start + PAGE_CACHE_SIZE - 1;
C
Chris Mason 已提交
7520

7521 7522 7523 7524 7525 7526 7527
	/*
	 * we have the page locked, so new writeback can't start,
	 * and the dirty bit won't be cleared while we are here.
	 *
	 * Wait for IO on this page so that we can safely clear
	 * the PagePrivate2 bit and do ordered accounting
	 */
7528
	wait_on_page_writeback(page);
7529

7530
	tree = &BTRFS_I(inode)->io_tree;
7531 7532 7533 7534
	if (offset) {
		btrfs_releasepage(page, GFP_NOFS);
		return;
	}
7535
	lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
M
Miao Xie 已提交
7536
	ordered = btrfs_lookup_ordered_extent(inode, page_offset(page));
7537
	if (ordered) {
7538 7539 7540 7541
		/*
		 * IO on this page will never be started, so we need
		 * to account for any ordered extents now
		 */
7542 7543
		clear_extent_bit(tree, page_start, page_end,
				 EXTENT_DIRTY | EXTENT_DELALLOC |
7544 7545
				 EXTENT_LOCKED | EXTENT_DO_ACCOUNTING |
				 EXTENT_DEFRAG, 1, 0, &cached_state, GFP_NOFS);
7546 7547 7548 7549
		/*
		 * whoever cleared the private bit is responsible
		 * for the finish_ordered_io
		 */
7550 7551 7552 7553
		if (TestClearPagePrivate2(page) &&
		    btrfs_dec_test_ordered_pending(inode, &ordered, page_start,
						   PAGE_CACHE_SIZE, 1)) {
			btrfs_finish_ordered_io(ordered);
7554
		}
7555
		btrfs_put_ordered_extent(ordered);
7556
		cached_state = NULL;
7557
		lock_extent_bits(tree, page_start, page_end, 0, &cached_state);
7558 7559
	}
	clear_extent_bit(tree, page_start, page_end,
7560
		 EXTENT_LOCKED | EXTENT_DIRTY | EXTENT_DELALLOC |
7561 7562
		 EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG, 1, 1,
		 &cached_state, GFP_NOFS);
7563 7564
	__btrfs_releasepage(page, GFP_NOFS);

C
Chris Mason 已提交
7565
	ClearPageChecked(page);
7566 7567 7568 7569 7570
	if (PagePrivate(page)) {
		ClearPagePrivate(page);
		set_page_private(page, 0);
		page_cache_release(page);
	}
C
Chris Mason 已提交
7571 7572
}

C
Chris Mason 已提交
7573 7574 7575 7576 7577 7578 7579 7580 7581 7582 7583 7584 7585 7586 7587
/*
 * btrfs_page_mkwrite() is not allowed to change the file size as it gets
 * called from a page fault handler when a page is first dirtied. Hence we must
 * be careful to check for EOF conditions here. We set the page up correctly
 * for a written page which means we get ENOSPC checking when writing into
 * holes and correct delalloc and unwritten extent mapping on filesystems that
 * support these features.
 *
 * We are not allowed to take the i_mutex here so we have to play games to
 * protect against truncate races as the page could now be beyond EOF.  Because
 * vmtruncate() writes the inode size before removing pages, once we have the
 * page lock we can determine safely if the page is beyond EOF. If it is not
 * beyond EOF, then the page is guaranteed safe against truncation until we
 * unlock the page.
 */
7588
int btrfs_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
C
Chris Mason 已提交
7589
{
7590
	struct page *page = vmf->page;
A
Al Viro 已提交
7591
	struct inode *inode = file_inode(vma->vm_file);
7592
	struct btrfs_root *root = BTRFS_I(inode)->root;
7593 7594
	struct extent_io_tree *io_tree = &BTRFS_I(inode)->io_tree;
	struct btrfs_ordered_extent *ordered;
7595
	struct extent_state *cached_state = NULL;
7596 7597
	char *kaddr;
	unsigned long zero_start;
C
Chris Mason 已提交
7598
	loff_t size;
7599
	int ret;
7600
	int reserved = 0;
7601
	u64 page_start;
7602
	u64 page_end;
C
Chris Mason 已提交
7603

7604
	sb_start_pagefault(inode->i_sb);
7605
	ret  = btrfs_delalloc_reserve_space(inode, PAGE_CACHE_SIZE);
7606
	if (!ret) {
7607
		ret = file_update_time(vma->vm_file);
7608 7609
		reserved = 1;
	}
7610 7611 7612 7613 7614
	if (ret) {
		if (ret == -ENOMEM)
			ret = VM_FAULT_OOM;
		else /* -ENOSPC, -EIO, etc */
			ret = VM_FAULT_SIGBUS;
7615 7616 7617
		if (reserved)
			goto out;
		goto out_noreserve;
7618
	}
7619

7620
	ret = VM_FAULT_NOPAGE; /* make the VM retry the fault */
7621
again:
C
Chris Mason 已提交
7622 7623
	lock_page(page);
	size = i_size_read(inode);
7624 7625
	page_start = page_offset(page);
	page_end = page_start + PAGE_CACHE_SIZE - 1;
7626

C
Chris Mason 已提交
7627
	if ((page->mapping != inode->i_mapping) ||
7628
	    (page_start >= size)) {
C
Chris Mason 已提交
7629 7630 7631
		/* page got truncated out from underneath us */
		goto out_unlock;
	}
7632 7633
	wait_on_page_writeback(page);

7634
	lock_extent_bits(io_tree, page_start, page_end, 0, &cached_state);
7635 7636
	set_page_extent_mapped(page);

7637 7638 7639 7640
	/*
	 * we can't set the delalloc bits if there are pending ordered
	 * extents.  Drop our locks and wait for them to finish
	 */
7641 7642
	ordered = btrfs_lookup_ordered_extent(inode, page_start);
	if (ordered) {
7643 7644
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
7645
		unlock_page(page);
7646
		btrfs_start_ordered_extent(inode, ordered, 1);
7647 7648 7649 7650
		btrfs_put_ordered_extent(ordered);
		goto again;
	}

J
Josef Bacik 已提交
7651 7652 7653 7654 7655 7656 7657
	/*
	 * XXX - page_mkwrite gets called every time the page is dirtied, even
	 * if it was already dirty, so for space accounting reasons we need to
	 * clear any delalloc bits for the range we are fixing to save.  There
	 * is probably a better way to do this, but for now keep consistent with
	 * prepare_pages in the normal write path.
	 */
7658
	clear_extent_bit(&BTRFS_I(inode)->io_tree, page_start, page_end,
7659 7660
			  EXTENT_DIRTY | EXTENT_DELALLOC |
			  EXTENT_DO_ACCOUNTING | EXTENT_DEFRAG,
7661
			  0, 0, &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
7662

7663 7664
	ret = btrfs_set_extent_delalloc(inode, page_start, page_end,
					&cached_state);
J
Josef Bacik 已提交
7665
	if (ret) {
7666 7667
		unlock_extent_cached(io_tree, page_start, page_end,
				     &cached_state, GFP_NOFS);
J
Josef Bacik 已提交
7668 7669 7670
		ret = VM_FAULT_SIGBUS;
		goto out_unlock;
	}
7671
	ret = 0;
C
Chris Mason 已提交
7672 7673

	/* page is wholly or partially inside EOF */
7674
	if (page_start + PAGE_CACHE_SIZE > size)
7675
		zero_start = size & ~PAGE_CACHE_MASK;
C
Chris Mason 已提交
7676
	else
7677
		zero_start = PAGE_CACHE_SIZE;
C
Chris Mason 已提交
7678

7679 7680 7681 7682 7683 7684
	if (zero_start != PAGE_CACHE_SIZE) {
		kaddr = kmap(page);
		memset(kaddr + zero_start, 0, PAGE_CACHE_SIZE - zero_start);
		flush_dcache_page(page);
		kunmap(page);
	}
7685
	ClearPageChecked(page);
7686
	set_page_dirty(page);
7687
	SetPageUptodate(page);
7688

7689 7690
	BTRFS_I(inode)->last_trans = root->fs_info->generation;
	BTRFS_I(inode)->last_sub_trans = BTRFS_I(inode)->root->log_transid;
7691
	BTRFS_I(inode)->last_log_commit = BTRFS_I(inode)->root->last_log_commit;
7692

7693
	unlock_extent_cached(io_tree, page_start, page_end, &cached_state, GFP_NOFS);
C
Chris Mason 已提交
7694 7695

out_unlock:
7696 7697
	if (!ret) {
		sb_end_pagefault(inode->i_sb);
7698
		return VM_FAULT_LOCKED;
7699
	}
C
Chris Mason 已提交
7700
	unlock_page(page);
7701
out:
7702
	btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
7703
out_noreserve:
7704
	sb_end_pagefault(inode->i_sb);
C
Chris Mason 已提交
7705 7706 7707
	return ret;
}

7708
static int btrfs_truncate(struct inode *inode)
C
Chris Mason 已提交
7709 7710
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
7711
	struct btrfs_block_rsv *rsv;
C
Chris Mason 已提交
7712
	int ret;
7713
	int err = 0;
C
Chris Mason 已提交
7714
	struct btrfs_trans_handle *trans;
7715
	u64 mask = root->sectorsize - 1;
7716
	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
C
Chris Mason 已提交
7717

J
Josef Bacik 已提交
7718
	ret = btrfs_truncate_page(inode, inode->i_size, 0, 0);
7719
	if (ret)
7720
		return ret;
7721

C
Chris Mason 已提交
7722
	btrfs_wait_ordered_range(inode, inode->i_size & (~mask), (u64)-1);
7723
	btrfs_ordered_update_i_size(inode, inode->i_size, NULL);
C
Chris Mason 已提交
7724

7725 7726 7727 7728 7729 7730 7731 7732 7733 7734 7735 7736 7737 7738 7739 7740 7741 7742 7743 7744 7745 7746 7747 7748 7749 7750 7751 7752 7753 7754 7755 7756 7757 7758 7759 7760
	/*
	 * Yes ladies and gentelment, this is indeed ugly.  The fact is we have
	 * 3 things going on here
	 *
	 * 1) We need to reserve space for our orphan item and the space to
	 * delete our orphan item.  Lord knows we don't want to have a dangling
	 * orphan item because we didn't reserve space to remove it.
	 *
	 * 2) We need to reserve space to update our inode.
	 *
	 * 3) We need to have something to cache all the space that is going to
	 * be free'd up by the truncate operation, but also have some slack
	 * space reserved in case it uses space during the truncate (thank you
	 * very much snapshotting).
	 *
	 * And we need these to all be seperate.  The fact is we can use alot of
	 * space doing the truncate, and we have no earthly idea how much space
	 * we will use, so we need the truncate reservation to be seperate so it
	 * doesn't end up using space reserved for updating the inode or
	 * removing the orphan item.  We also need to be able to stop the
	 * transaction and start a new one, which means we need to be able to
	 * update the inode several times, and we have no idea of knowing how
	 * many times that will be, so we can't just reserve 1 item for the
	 * entirety of the opration, so that has to be done seperately as well.
	 * Then there is the orphan item, which does indeed need to be held on
	 * to for the whole operation, and we need nobody to touch this reserved
	 * space except the orphan code.
	 *
	 * So that leaves us with
	 *
	 * 1) root->orphan_block_rsv - for the orphan deletion.
	 * 2) rsv - for the truncate reservation, which we will steal from the
	 * transaction reservation.
	 * 3) fs_info->trans_block_rsv - this will have 1 items worth left for
	 * updating the inode.
	 */
7761
	rsv = btrfs_alloc_block_rsv(root, BTRFS_BLOCK_RSV_TEMP);
7762 7763
	if (!rsv)
		return -ENOMEM;
J
Josef Bacik 已提交
7764
	rsv->size = min_size;
7765
	rsv->failfast = 1;
7766

7767
	/*
7768
	 * 1 for the truncate slack space
7769 7770
	 * 1 for updating the inode.
	 */
7771
	trans = btrfs_start_transaction(root, 2);
7772 7773 7774 7775
	if (IS_ERR(trans)) {
		err = PTR_ERR(trans);
		goto out;
	}
7776

7777 7778 7779
	/* Migrate the slack space for the truncate to our reserve */
	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
				      min_size);
7780
	BUG_ON(ret);
7781

7782 7783 7784 7785 7786 7787 7788 7789 7790 7791 7792 7793 7794 7795 7796 7797 7798
	/*
	 * setattr is responsible for setting the ordered_data_close flag,
	 * but that is only tested during the last file release.  That
	 * could happen well after the next commit, leaving a great big
	 * window where new writes may get lost if someone chooses to write
	 * to this file after truncating to zero
	 *
	 * The inode doesn't have any dirty data here, and so if we commit
	 * this is a noop.  If someone immediately starts writing to the inode
	 * it is very likely we'll catch some of their writes in this
	 * transaction, and the commit will find this file on the ordered
	 * data list with good things to send down.
	 *
	 * This is a best effort solution, there is still a window where
	 * using truncate to replace the contents of the file will
	 * end up with a zero length file after a crash.
	 */
7799 7800
	if (inode->i_size == 0 && test_bit(BTRFS_INODE_ORDERED_DATA_CLOSE,
					   &BTRFS_I(inode)->runtime_flags))
7801 7802
		btrfs_add_ordered_operation(trans, root, inode);

J
Josef Bacik 已提交
7803 7804 7805 7806 7807 7808 7809 7810
	/*
	 * So if we truncate and then write and fsync we normally would just
	 * write the extents that changed, which is a problem if we need to
	 * first truncate that entire inode.  So set this flag so we write out
	 * all of the extents in the inode to the sync log so we're completely
	 * safe.
	 */
	set_bit(BTRFS_INODE_NEEDS_FULL_SYNC, &BTRFS_I(inode)->runtime_flags);
7811
	trans->block_rsv = rsv;
7812

7813 7814 7815 7816
	while (1) {
		ret = btrfs_truncate_inode_items(trans, root, inode,
						 inode->i_size,
						 BTRFS_EXTENT_DATA_KEY);
7817
		if (ret != -ENOSPC) {
7818
			err = ret;
7819
			break;
7820
		}
C
Chris Mason 已提交
7821

7822
		trans->block_rsv = &root->fs_info->trans_block_rsv;
7823
		ret = btrfs_update_inode(trans, root, inode);
7824 7825 7826 7827
		if (ret) {
			err = ret;
			break;
		}
7828

7829
		btrfs_end_transaction(trans, root);
7830
		btrfs_btree_balance_dirty(root);
7831 7832 7833 7834 7835 7836 7837 7838 7839 7840 7841 7842

		trans = btrfs_start_transaction(root, 2);
		if (IS_ERR(trans)) {
			ret = err = PTR_ERR(trans);
			trans = NULL;
			break;
		}

		ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv,
					      rsv, min_size);
		BUG_ON(ret);	/* shouldn't happen */
		trans->block_rsv = rsv;
7843 7844 7845
	}

	if (ret == 0 && inode->i_nlink > 0) {
7846
		trans->block_rsv = root->orphan_block_rsv;
7847
		ret = btrfs_orphan_del(trans, inode);
7848 7849
		if (ret)
			err = ret;
7850 7851
	}

7852 7853 7854 7855 7856
	if (trans) {
		trans->block_rsv = &root->fs_info->trans_block_rsv;
		ret = btrfs_update_inode(trans, root, inode);
		if (ret && !err)
			err = ret;
7857

7858
		ret = btrfs_end_transaction(trans, root);
7859
		btrfs_btree_balance_dirty(root);
7860
	}
7861 7862 7863 7864

out:
	btrfs_free_block_rsv(root, rsv);

7865 7866
	if (ret && !err)
		err = ret;
7867

7868
	return err;
C
Chris Mason 已提交
7869 7870
}

C
Chris Mason 已提交
7871 7872 7873
/*
 * create a new subvolume directory/inode (helper for the ioctl).
 */
7874
int btrfs_create_subvol_root(struct btrfs_trans_handle *trans,
7875
			     struct btrfs_root *new_root, u64 new_dirid)
C
Chris Mason 已提交
7876 7877
{
	struct inode *inode;
7878
	int err;
7879
	u64 index = 0;
C
Chris Mason 已提交
7880

7881 7882 7883 7884
	inode = btrfs_new_inode(trans, new_root, NULL, "..", 2,
				new_dirid, new_dirid,
				S_IFDIR | (~current_umask() & S_IRWXUGO),
				&index);
7885
	if (IS_ERR(inode))
C
Christoph Hellwig 已提交
7886
		return PTR_ERR(inode);
C
Chris Mason 已提交
7887 7888 7889
	inode->i_op = &btrfs_dir_inode_operations;
	inode->i_fop = &btrfs_dir_file_operations;

M
Miklos Szeredi 已提交
7890
	set_nlink(inode, 1);
7891
	btrfs_i_size_write(inode, 0);
7892

7893
	err = btrfs_update_inode(trans, new_root, inode);
7894

7895
	iput(inode);
7896
	return err;
C
Chris Mason 已提交
7897 7898 7899 7900 7901
}

struct inode *btrfs_alloc_inode(struct super_block *sb)
{
	struct btrfs_inode *ei;
Y
Yan, Zheng 已提交
7902
	struct inode *inode;
C
Chris Mason 已提交
7903 7904 7905 7906

	ei = kmem_cache_alloc(btrfs_inode_cachep, GFP_NOFS);
	if (!ei)
		return NULL;
Y
Yan, Zheng 已提交
7907 7908 7909

	ei->root = NULL;
	ei->generation = 0;
7910
	ei->last_trans = 0;
7911
	ei->last_sub_trans = 0;
7912
	ei->logged_trans = 0;
Y
Yan, Zheng 已提交
7913 7914 7915
	ei->delalloc_bytes = 0;
	ei->disk_i_size = 0;
	ei->flags = 0;
7916
	ei->csum_bytes = 0;
Y
Yan, Zheng 已提交
7917 7918
	ei->index_cnt = (u64)-1;
	ei->last_unlink_trans = 0;
7919
	ei->last_log_commit = 0;
Y
Yan, Zheng 已提交
7920

7921 7922 7923
	spin_lock_init(&ei->lock);
	ei->outstanding_extents = 0;
	ei->reserved_extents = 0;
Y
Yan, Zheng 已提交
7924

7925
	ei->runtime_flags = 0;
7926
	ei->force_compress = BTRFS_COMPRESS_NONE;
Y
Yan, Zheng 已提交
7927

7928 7929
	ei->delayed_node = NULL;

Y
Yan, Zheng 已提交
7930
	inode = &ei->vfs_inode;
7931
	extent_map_tree_init(&ei->extent_tree);
7932 7933
	extent_io_tree_init(&ei->io_tree, &inode->i_data);
	extent_io_tree_init(&ei->io_failure_tree, &inode->i_data);
7934 7935
	ei->io_tree.track_uptodate = 1;
	ei->io_failure_tree.track_uptodate = 1;
7936
	atomic_set(&ei->sync_writers, 0);
Y
Yan, Zheng 已提交
7937
	mutex_init(&ei->log_mutex);
7938
	mutex_init(&ei->delalloc_mutex);
7939
	btrfs_ordered_inode_tree_init(&ei->ordered_tree);
Y
Yan, Zheng 已提交
7940
	INIT_LIST_HEAD(&ei->delalloc_inodes);
7941
	INIT_LIST_HEAD(&ei->ordered_operations);
Y
Yan, Zheng 已提交
7942 7943 7944
	RB_CLEAR_NODE(&ei->rb_node);

	return inode;
C
Chris Mason 已提交
7945 7946
}

N
Nick Piggin 已提交
7947 7948 7949 7950 7951 7952
static void btrfs_i_callback(struct rcu_head *head)
{
	struct inode *inode = container_of(head, struct inode, i_rcu);
	kmem_cache_free(btrfs_inode_cachep, BTRFS_I(inode));
}

C
Chris Mason 已提交
7953 7954
void btrfs_destroy_inode(struct inode *inode)
{
7955
	struct btrfs_ordered_extent *ordered;
7956 7957
	struct btrfs_root *root = BTRFS_I(inode)->root;

A
Al Viro 已提交
7958
	WARN_ON(!hlist_empty(&inode->i_dentry));
C
Chris Mason 已提交
7959
	WARN_ON(inode->i_data.nrpages);
7960 7961
	WARN_ON(BTRFS_I(inode)->outstanding_extents);
	WARN_ON(BTRFS_I(inode)->reserved_extents);
7962 7963
	WARN_ON(BTRFS_I(inode)->delalloc_bytes);
	WARN_ON(BTRFS_I(inode)->csum_bytes);
C
Chris Mason 已提交
7964

7965 7966 7967 7968 7969 7970 7971 7972
	/*
	 * This can happen where we create an inode, but somebody else also
	 * created the same inode and we need to destroy the one we already
	 * created.
	 */
	if (!root)
		goto free;

7973 7974 7975 7976 7977 7978 7979 7980 7981 7982 7983
	/*
	 * Make sure we're properly removed from the ordered operation
	 * lists.
	 */
	smp_mb();
	if (!list_empty(&BTRFS_I(inode)->ordered_operations)) {
		spin_lock(&root->fs_info->ordered_extent_lock);
		list_del_init(&BTRFS_I(inode)->ordered_operations);
		spin_unlock(&root->fs_info->ordered_extent_lock);
	}

7984 7985
	if (test_bit(BTRFS_INODE_HAS_ORPHAN_ITEM,
		     &BTRFS_I(inode)->runtime_flags)) {
7986 7987
		btrfs_info(root->fs_info, "inode %llu still on the orphan list",
			(unsigned long long)btrfs_ino(inode));
7988
		atomic_dec(&root->orphan_inodes);
7989 7990
	}

C
Chris Mason 已提交
7991
	while (1) {
7992 7993 7994 7995
		ordered = btrfs_lookup_first_ordered_extent(inode, (u64)-1);
		if (!ordered)
			break;
		else {
7996 7997 7998
			btrfs_err(root->fs_info, "found ordered extent %llu %llu on inode cleanup",
				(unsigned long long)ordered->file_offset,
				(unsigned long long)ordered->len);
7999 8000 8001 8002 8003
			btrfs_remove_ordered_extent(inode, ordered);
			btrfs_put_ordered_extent(ordered);
			btrfs_put_ordered_extent(ordered);
		}
	}
8004
	inode_tree_del(inode);
8005
	btrfs_drop_extent_cache(inode, 0, (u64)-1, 0);
8006
free:
N
Nick Piggin 已提交
8007
	call_rcu(&inode->i_rcu, btrfs_i_callback);
C
Chris Mason 已提交
8008 8009
}

8010
int btrfs_drop_inode(struct inode *inode)
8011 8012
{
	struct btrfs_root *root = BTRFS_I(inode)->root;
8013

8014 8015 8016
	if (root == NULL)
		return 1;

8017
	/* the snap/subvol tree is on deleting */
8018
	if (btrfs_root_refs(&root->root_item) == 0 &&
8019
	    root != root->fs_info->tree_root)
8020
		return 1;
8021
	else
8022
		return generic_drop_inode(inode);
8023 8024
}

8025
static void init_once(void *foo)
C
Chris Mason 已提交
8026 8027 8028 8029 8030 8031 8032 8033
{
	struct btrfs_inode *ei = (struct btrfs_inode *) foo;

	inode_init_once(&ei->vfs_inode);
}

void btrfs_destroy_cachep(void)
{
8034 8035 8036 8037 8038
	/*
	 * Make sure all delayed rcu free inodes are flushed before we
	 * destroy cache.
	 */
	rcu_barrier();
C
Chris Mason 已提交
8039 8040 8041 8042 8043 8044 8045 8046
	if (btrfs_inode_cachep)
		kmem_cache_destroy(btrfs_inode_cachep);
	if (btrfs_trans_handle_cachep)
		kmem_cache_destroy(btrfs_trans_handle_cachep);
	if (btrfs_transaction_cachep)
		kmem_cache_destroy(btrfs_transaction_cachep);
	if (btrfs_path_cachep)
		kmem_cache_destroy(btrfs_path_cachep);
8047 8048
	if (btrfs_free_space_cachep)
		kmem_cache_destroy(btrfs_free_space_cachep);
8049 8050
	if (btrfs_delalloc_work_cachep)
		kmem_cache_destroy(btrfs_delalloc_work_cachep);
C
Chris Mason 已提交
8051 8052 8053 8054
}

int btrfs_init_cachep(void)
{
D
David Sterba 已提交
8055
	btrfs_inode_cachep = kmem_cache_create("btrfs_inode",
8056 8057
			sizeof(struct btrfs_inode), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, init_once);
C
Chris Mason 已提交
8058 8059
	if (!btrfs_inode_cachep)
		goto fail;
8060

D
David Sterba 已提交
8061
	btrfs_trans_handle_cachep = kmem_cache_create("btrfs_trans_handle",
8062 8063
			sizeof(struct btrfs_trans_handle), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
8064 8065
	if (!btrfs_trans_handle_cachep)
		goto fail;
8066

D
David Sterba 已提交
8067
	btrfs_transaction_cachep = kmem_cache_create("btrfs_transaction",
8068 8069
			sizeof(struct btrfs_transaction), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
8070 8071
	if (!btrfs_transaction_cachep)
		goto fail;
8072

D
David Sterba 已提交
8073
	btrfs_path_cachep = kmem_cache_create("btrfs_path",
8074 8075
			sizeof(struct btrfs_path), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
C
Chris Mason 已提交
8076 8077
	if (!btrfs_path_cachep)
		goto fail;
8078

D
David Sterba 已提交
8079
	btrfs_free_space_cachep = kmem_cache_create("btrfs_free_space",
8080 8081 8082 8083 8084
			sizeof(struct btrfs_free_space), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, NULL);
	if (!btrfs_free_space_cachep)
		goto fail;

8085 8086 8087 8088 8089 8090 8091
	btrfs_delalloc_work_cachep = kmem_cache_create("btrfs_delalloc_work",
			sizeof(struct btrfs_delalloc_work), 0,
			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
			NULL);
	if (!btrfs_delalloc_work_cachep)
		goto fail;

C
Chris Mason 已提交
8092 8093 8094 8095 8096 8097 8098 8099 8100
	return 0;
fail:
	btrfs_destroy_cachep();
	return -ENOMEM;
}

static int btrfs_getattr(struct vfsmount *mnt,
			 struct dentry *dentry, struct kstat *stat)
{
8101
	u64 delalloc_bytes;
C
Chris Mason 已提交
8102
	struct inode *inode = dentry->d_inode;
D
David Sterba 已提交
8103 8104
	u32 blocksize = inode->i_sb->s_blocksize;

C
Chris Mason 已提交
8105
	generic_fillattr(inode, stat);
8106
	stat->dev = BTRFS_I(inode)->root->anon_dev;
C
Chris Mason 已提交
8107
	stat->blksize = PAGE_CACHE_SIZE;
8108 8109 8110 8111

	spin_lock(&BTRFS_I(inode)->lock);
	delalloc_bytes = BTRFS_I(inode)->delalloc_bytes;
	spin_unlock(&BTRFS_I(inode)->lock);
D
David Sterba 已提交
8112
	stat->blocks = (ALIGN(inode_get_bytes(inode), blocksize) +
8113
			ALIGN(delalloc_bytes, blocksize)) >> 9;
C
Chris Mason 已提交
8114 8115 8116
	return 0;
}

C
Chris Mason 已提交
8117 8118
static int btrfs_rename(struct inode *old_dir, struct dentry *old_dentry,
			   struct inode *new_dir, struct dentry *new_dentry)
C
Chris Mason 已提交
8119 8120 8121
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(old_dir)->root;
8122
	struct btrfs_root *dest = BTRFS_I(new_dir)->root;
C
Chris Mason 已提交
8123 8124 8125
	struct inode *new_inode = new_dentry->d_inode;
	struct inode *old_inode = old_dentry->d_inode;
	struct timespec ctime = CURRENT_TIME;
8126
	u64 index = 0;
8127
	u64 root_objectid;
C
Chris Mason 已提交
8128
	int ret;
L
Li Zefan 已提交
8129
	u64 old_ino = btrfs_ino(old_inode);
C
Chris Mason 已提交
8130

L
Li Zefan 已提交
8131
	if (btrfs_ino(new_dir) == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)
8132 8133
		return -EPERM;

8134
	/* we only allow rename subvolume link between subvolumes */
L
Li Zefan 已提交
8135
	if (old_ino != BTRFS_FIRST_FREE_OBJECTID && root != dest)
8136 8137
		return -EXDEV;

L
Li Zefan 已提交
8138 8139
	if (old_ino == BTRFS_EMPTY_SUBVOL_DIR_OBJECTID ||
	    (new_inode && btrfs_ino(new_inode) == BTRFS_FIRST_FREE_OBJECTID))
C
Chris Mason 已提交
8140
		return -ENOTEMPTY;
8141

8142 8143 8144
	if (S_ISDIR(old_inode->i_mode) && new_inode &&
	    new_inode->i_size > BTRFS_EMPTY_DIR_SIZE)
		return -ENOTEMPTY;
C
Chris Mason 已提交
8145 8146 8147 8148 8149 8150 8151 8152 8153 8154 8155 8156 8157 8158 8159 8160 8161 8162 8163 8164 8165 8166


	/* check for collisions, even if the  name isn't there */
	ret = btrfs_check_dir_item_collision(root, new_dir->i_ino,
			     new_dentry->d_name.name,
			     new_dentry->d_name.len);

	if (ret) {
		if (ret == -EEXIST) {
			/* we shouldn't get
			 * eexist without a new_inode */
			if (!new_inode) {
				WARN_ON(1);
				return ret;
			}
		} else {
			/* maybe -EOVERFLOW */
			return ret;
		}
	}
	ret = 0;

8167 8168 8169 8170 8171
	/*
	 * we're using rename to replace one file with another.
	 * and the replacement file is large.  Start IO on it now so
	 * we don't add too much work to the end of the transaction
	 */
8172
	if (new_inode && S_ISREG(old_inode->i_mode) && new_inode->i_size &&
8173 8174 8175
	    old_inode->i_size > BTRFS_ORDERED_OPERATIONS_FLUSH_LIMIT)
		filemap_flush(old_inode->i_mapping);

8176
	/* close the racy window with snapshot create/destroy ioctl */
L
Li Zefan 已提交
8177
	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
8178
		down_read(&root->fs_info->subvol_sem);
8179 8180 8181 8182 8183 8184 8185 8186
	/*
	 * We want to reserve the absolute worst case amount of items.  So if
	 * both inodes are subvols and we need to unlink them then that would
	 * require 4 item modifications, but if they are both normal inodes it
	 * would require 5 item modifications, so we'll assume their normal
	 * inodes.  So 5 * 2 is 10, plus 1 for the new link, so 11 total items
	 * should cover the worst case number of items we'll modify.
	 */
8187
	trans = btrfs_start_transaction(root, 11);
8188 8189 8190 8191
	if (IS_ERR(trans)) {
                ret = PTR_ERR(trans);
                goto out_notrans;
        }
8192

8193 8194
	if (dest != root)
		btrfs_record_root_in_trans(trans, dest);
8195

8196 8197 8198
	ret = btrfs_set_inode_index(new_dir, &index);
	if (ret)
		goto out_fail;
8199

L
Li Zefan 已提交
8200
	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
8201 8202 8203
		/* force full log commit if subvolume involved. */
		root->fs_info->last_trans_log_full_commit = trans->transid;
	} else {
8204 8205 8206
		ret = btrfs_insert_inode_ref(trans, dest,
					     new_dentry->d_name.name,
					     new_dentry->d_name.len,
L
Li Zefan 已提交
8207 8208
					     old_ino,
					     btrfs_ino(new_dir), index);
8209 8210
		if (ret)
			goto out_fail;
8211 8212 8213 8214 8215 8216 8217 8218 8219
		/*
		 * this is an ugly little race, but the rename is required
		 * to make sure that if we crash, the inode is either at the
		 * old name or the new one.  pinning the log transaction lets
		 * us make sure we don't allow a log commit to come in after
		 * we unlink the name but before we add the new name back in.
		 */
		btrfs_pin_log_trans(root);
	}
8220 8221 8222 8223
	/*
	 * make sure the inode gets flushed if it is replacing
	 * something.
	 */
L
Li Zefan 已提交
8224
	if (new_inode && new_inode->i_size && S_ISREG(old_inode->i_mode))
8225 8226
		btrfs_add_ordered_operation(trans, root, old_inode);

8227 8228 8229
	inode_inc_iversion(old_dir);
	inode_inc_iversion(new_dir);
	inode_inc_iversion(old_inode);
C
Chris Mason 已提交
8230 8231 8232
	old_dir->i_ctime = old_dir->i_mtime = ctime;
	new_dir->i_ctime = new_dir->i_mtime = ctime;
	old_inode->i_ctime = ctime;
8233

8234 8235 8236
	if (old_dentry->d_parent != new_dentry->d_parent)
		btrfs_record_unlink_dir(trans, old_dir, old_inode, 1);

L
Li Zefan 已提交
8237
	if (unlikely(old_ino == BTRFS_FIRST_FREE_OBJECTID)) {
8238 8239 8240 8241 8242
		root_objectid = BTRFS_I(old_inode)->root->root_key.objectid;
		ret = btrfs_unlink_subvol(trans, root, old_dir, root_objectid,
					old_dentry->d_name.name,
					old_dentry->d_name.len);
	} else {
8243 8244 8245 8246 8247 8248
		ret = __btrfs_unlink_inode(trans, root, old_dir,
					old_dentry->d_inode,
					old_dentry->d_name.name,
					old_dentry->d_name.len);
		if (!ret)
			ret = btrfs_update_inode(trans, root, old_inode);
8249
	}
8250 8251 8252 8253
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_fail;
	}
C
Chris Mason 已提交
8254 8255

	if (new_inode) {
8256
		inode_inc_iversion(new_inode);
C
Chris Mason 已提交
8257
		new_inode->i_ctime = CURRENT_TIME;
L
Li Zefan 已提交
8258
		if (unlikely(btrfs_ino(new_inode) ==
8259 8260 8261 8262 8263 8264 8265 8266 8267 8268 8269 8270 8271
			     BTRFS_EMPTY_SUBVOL_DIR_OBJECTID)) {
			root_objectid = BTRFS_I(new_inode)->location.objectid;
			ret = btrfs_unlink_subvol(trans, dest, new_dir,
						root_objectid,
						new_dentry->d_name.name,
						new_dentry->d_name.len);
			BUG_ON(new_inode->i_nlink == 0);
		} else {
			ret = btrfs_unlink_inode(trans, dest, new_dir,
						 new_dentry->d_inode,
						 new_dentry->d_name.name,
						 new_dentry->d_name.len);
		}
8272
		if (!ret && new_inode->i_nlink == 0) {
8273
			ret = btrfs_orphan_add(trans, new_dentry->d_inode);
8274
			BUG_ON(ret);
8275
		}
8276 8277 8278 8279
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			goto out_fail;
		}
C
Chris Mason 已提交
8280
	}
8281

8282 8283
	ret = btrfs_add_link(trans, new_dir, old_inode,
			     new_dentry->d_name.name,
8284
			     new_dentry->d_name.len, 0, index);
8285 8286 8287 8288
	if (ret) {
		btrfs_abort_transaction(trans, root, ret);
		goto out_fail;
	}
C
Chris Mason 已提交
8289

L
Li Zefan 已提交
8290
	if (old_ino != BTRFS_FIRST_FREE_OBJECTID) {
8291
		struct dentry *parent = new_dentry->d_parent;
8292
		btrfs_log_new_name(trans, old_inode, old_dir, parent);
8293 8294
		btrfs_end_log_trans(root);
	}
C
Chris Mason 已提交
8295
out_fail:
8296
	btrfs_end_transaction(trans, root);
8297
out_notrans:
L
Li Zefan 已提交
8298
	if (old_ino == BTRFS_FIRST_FREE_OBJECTID)
8299
		up_read(&root->fs_info->subvol_sem);
J
Josef Bacik 已提交
8300

C
Chris Mason 已提交
8301 8302 8303
	return ret;
}

8304 8305 8306 8307 8308 8309 8310 8311 8312 8313 8314 8315 8316 8317 8318 8319 8320 8321 8322 8323 8324 8325 8326 8327 8328 8329 8330 8331 8332 8333 8334 8335 8336 8337 8338 8339 8340 8341 8342 8343 8344 8345 8346
static void btrfs_run_delalloc_work(struct btrfs_work *work)
{
	struct btrfs_delalloc_work *delalloc_work;

	delalloc_work = container_of(work, struct btrfs_delalloc_work,
				     work);
	if (delalloc_work->wait)
		btrfs_wait_ordered_range(delalloc_work->inode, 0, (u64)-1);
	else
		filemap_flush(delalloc_work->inode->i_mapping);

	if (delalloc_work->delay_iput)
		btrfs_add_delayed_iput(delalloc_work->inode);
	else
		iput(delalloc_work->inode);
	complete(&delalloc_work->completion);
}

struct btrfs_delalloc_work *btrfs_alloc_delalloc_work(struct inode *inode,
						    int wait, int delay_iput)
{
	struct btrfs_delalloc_work *work;

	work = kmem_cache_zalloc(btrfs_delalloc_work_cachep, GFP_NOFS);
	if (!work)
		return NULL;

	init_completion(&work->completion);
	INIT_LIST_HEAD(&work->list);
	work->inode = inode;
	work->wait = wait;
	work->delay_iput = delay_iput;
	work->work.func = btrfs_run_delalloc_work;

	return work;
}

void btrfs_wait_and_free_delalloc_work(struct btrfs_delalloc_work *work)
{
	wait_for_completion(&work->completion);
	kmem_cache_free(btrfs_delalloc_work_cachep, work);
}

C
Chris Mason 已提交
8347 8348 8349 8350
/*
 * some fairly slow code that needs optimization. This walks the list
 * of all the inodes with pending delalloc and forces them to disk.
 */
Y
Yan, Zheng 已提交
8351
int btrfs_start_delalloc_inodes(struct btrfs_root *root, int delay_iput)
8352 8353
{
	struct btrfs_inode *binode;
8354
	struct inode *inode;
8355 8356
	struct btrfs_delalloc_work *work, *next;
	struct list_head works;
8357
	struct list_head splice;
8358
	int ret = 0;
8359

Y
Yan Zheng 已提交
8360 8361 8362
	if (root->fs_info->sb->s_flags & MS_RDONLY)
		return -EROFS;

8363
	INIT_LIST_HEAD(&works);
8364
	INIT_LIST_HEAD(&splice);
8365

8366
	spin_lock(&root->fs_info->delalloc_lock);
8367 8368 8369
	list_splice_init(&root->fs_info->delalloc_inodes, &splice);
	while (!list_empty(&splice)) {
		binode = list_entry(splice.next, struct btrfs_inode,
8370
				    delalloc_inodes);
8371 8372 8373

		list_del_init(&binode->delalloc_inodes);

8374
		inode = igrab(&binode->vfs_inode);
8375 8376 8377
		if (!inode) {
			clear_bit(BTRFS_INODE_IN_DELALLOC_LIST,
				  &binode->runtime_flags);
8378
			continue;
8379
		}
8380 8381 8382

		list_add_tail(&binode->delalloc_inodes,
			      &root->fs_info->delalloc_inodes);
8383
		spin_unlock(&root->fs_info->delalloc_lock);
8384 8385 8386 8387 8388

		work = btrfs_alloc_delalloc_work(inode, 0, delay_iput);
		if (unlikely(!work)) {
			ret = -ENOMEM;
			goto out;
8389
		}
8390 8391 8392 8393
		list_add_tail(&work->list, &works);
		btrfs_queue_worker(&root->fs_info->flush_workers,
				   &work->work);

8394
		cond_resched();
8395
		spin_lock(&root->fs_info->delalloc_lock);
8396
	}
8397
	spin_unlock(&root->fs_info->delalloc_lock);
8398

8399 8400 8401 8402 8403
	list_for_each_entry_safe(work, next, &works, list) {
		list_del_init(&work->list);
		btrfs_wait_and_free_delalloc_work(work);
	}

8404 8405 8406 8407 8408
	/* the filemap_flush will queue IO into the worker threads, but
	 * we have to make sure the IO is actually started and that
	 * ordered extents get created before we return
	 */
	atomic_inc(&root->fs_info->async_submit_draining);
C
Chris Mason 已提交
8409
	while (atomic_read(&root->fs_info->nr_async_submits) ||
8410
	      atomic_read(&root->fs_info->async_delalloc_pages)) {
8411
		wait_event(root->fs_info->async_submit_wait,
8412 8413
		   (atomic_read(&root->fs_info->nr_async_submits) == 0 &&
		    atomic_read(&root->fs_info->async_delalloc_pages) == 0));
8414 8415
	}
	atomic_dec(&root->fs_info->async_submit_draining);
8416
	return 0;
8417 8418 8419 8420 8421
out:
	list_for_each_entry_safe(work, next, &works, list) {
		list_del_init(&work->list);
		btrfs_wait_and_free_delalloc_work(work);
	}
8422 8423 8424 8425 8426 8427

	if (!list_empty_careful(&splice)) {
		spin_lock(&root->fs_info->delalloc_lock);
		list_splice_tail(&splice, &root->fs_info->delalloc_inodes);
		spin_unlock(&root->fs_info->delalloc_lock);
	}
8428
	return ret;
8429 8430
}

C
Chris Mason 已提交
8431 8432 8433 8434 8435 8436 8437
static int btrfs_symlink(struct inode *dir, struct dentry *dentry,
			 const char *symname)
{
	struct btrfs_trans_handle *trans;
	struct btrfs_root *root = BTRFS_I(dir)->root;
	struct btrfs_path *path;
	struct btrfs_key key;
8438
	struct inode *inode = NULL;
C
Chris Mason 已提交
8439 8440 8441
	int err;
	int drop_inode = 0;
	u64 objectid;
8442
	u64 index = 0 ;
C
Chris Mason 已提交
8443 8444
	int name_len;
	int datasize;
8445
	unsigned long ptr;
C
Chris Mason 已提交
8446
	struct btrfs_file_extent_item *ei;
8447
	struct extent_buffer *leaf;
C
Chris Mason 已提交
8448 8449 8450 8451

	name_len = strlen(symname) + 1;
	if (name_len > BTRFS_MAX_INLINE_DATA_SIZE(root))
		return -ENAMETOOLONG;
8452

J
Josef Bacik 已提交
8453 8454 8455 8456 8457
	/*
	 * 2 items for inode item and ref
	 * 2 items for dir items
	 * 1 item for xattr if selinux is on
	 */
8458 8459 8460
	trans = btrfs_start_transaction(root, 5);
	if (IS_ERR(trans))
		return PTR_ERR(trans);
8461

8462 8463 8464 8465
	err = btrfs_find_free_ino(root, &objectid);
	if (err)
		goto out_unlock;

8466
	inode = btrfs_new_inode(trans, root, dir, dentry->d_name.name,
L
Li Zefan 已提交
8467
				dentry->d_name.len, btrfs_ino(dir), objectid,
8468
				S_IFLNK|S_IRWXUGO, &index);
8469 8470
	if (IS_ERR(inode)) {
		err = PTR_ERR(inode);
C
Chris Mason 已提交
8471
		goto out_unlock;
8472
	}
C
Chris Mason 已提交
8473

8474
	err = btrfs_init_inode_security(trans, inode, dir, &dentry->d_name);
J
Josef Bacik 已提交
8475 8476 8477 8478 8479
	if (err) {
		drop_inode = 1;
		goto out_unlock;
	}

8480 8481 8482 8483 8484 8485 8486 8487 8488
	/*
	* If the active LSM wants to access the inode during
	* d_instantiate it needs these. Smack checks to see
	* if the filesystem supports xattrs by looking at the
	* ops vector.
	*/
	inode->i_fop = &btrfs_file_operations;
	inode->i_op = &btrfs_file_inode_operations;

8489
	err = btrfs_add_nondir(trans, dir, dentry, inode, 0, index);
C
Chris Mason 已提交
8490 8491 8492 8493
	if (err)
		drop_inode = 1;
	else {
		inode->i_mapping->a_ops = &btrfs_aops;
C
Chris Mason 已提交
8494
		inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
8495
		BTRFS_I(inode)->io_tree.ops = &btrfs_extent_io_ops;
C
Chris Mason 已提交
8496 8497 8498 8499 8500
	}
	if (drop_inode)
		goto out_unlock;

	path = btrfs_alloc_path();
8501 8502 8503 8504 8505
	if (!path) {
		err = -ENOMEM;
		drop_inode = 1;
		goto out_unlock;
	}
L
Li Zefan 已提交
8506
	key.objectid = btrfs_ino(inode);
C
Chris Mason 已提交
8507 8508 8509 8510 8511
	key.offset = 0;
	btrfs_set_key_type(&key, BTRFS_EXTENT_DATA_KEY);
	datasize = btrfs_file_extent_calc_inline_size(name_len);
	err = btrfs_insert_empty_item(trans, root, path, &key,
				      datasize);
8512 8513
	if (err) {
		drop_inode = 1;
8514
		btrfs_free_path(path);
8515 8516
		goto out_unlock;
	}
8517 8518 8519 8520 8521
	leaf = path->nodes[0];
	ei = btrfs_item_ptr(leaf, path->slots[0],
			    struct btrfs_file_extent_item);
	btrfs_set_file_extent_generation(leaf, ei, trans->transid);
	btrfs_set_file_extent_type(leaf, ei,
C
Chris Mason 已提交
8522
				   BTRFS_FILE_EXTENT_INLINE);
C
Chris Mason 已提交
8523 8524 8525 8526 8527
	btrfs_set_file_extent_encryption(leaf, ei, 0);
	btrfs_set_file_extent_compression(leaf, ei, 0);
	btrfs_set_file_extent_other_encoding(leaf, ei, 0);
	btrfs_set_file_extent_ram_bytes(leaf, ei, name_len);

C
Chris Mason 已提交
8528
	ptr = btrfs_file_extent_inline_start(ei);
8529 8530
	write_extent_buffer(leaf, symname, ptr, name_len);
	btrfs_mark_buffer_dirty(leaf);
C
Chris Mason 已提交
8531
	btrfs_free_path(path);
8532

C
Chris Mason 已提交
8533 8534
	inode->i_op = &btrfs_symlink_inode_operations;
	inode->i_mapping->a_ops = &btrfs_symlink_aops;
C
Chris Mason 已提交
8535
	inode->i_mapping->backing_dev_info = &root->fs_info->bdi;
Y
Yan Zheng 已提交
8536
	inode_set_bytes(inode, name_len);
8537
	btrfs_i_size_write(inode, name_len - 1);
8538 8539 8540
	err = btrfs_update_inode(trans, root, inode);
	if (err)
		drop_inode = 1;
C
Chris Mason 已提交
8541 8542

out_unlock:
8543 8544
	if (!err)
		d_instantiate(dentry, inode);
8545
	btrfs_end_transaction(trans, root);
C
Chris Mason 已提交
8546 8547 8548 8549
	if (drop_inode) {
		inode_dec_link_count(inode);
		iput(inode);
	}
8550
	btrfs_btree_balance_dirty(root);
C
Chris Mason 已提交
8551 8552
	return err;
}
8553

8554 8555 8556 8557
static int __btrfs_prealloc_file_range(struct inode *inode, int mode,
				       u64 start, u64 num_bytes, u64 min_size,
				       loff_t actual_len, u64 *alloc_hint,
				       struct btrfs_trans_handle *trans)
Y
Yan Zheng 已提交
8558
{
J
Josef Bacik 已提交
8559 8560
	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
	struct extent_map *em;
Y
Yan Zheng 已提交
8561 8562 8563
	struct btrfs_root *root = BTRFS_I(inode)->root;
	struct btrfs_key ins;
	u64 cur_offset = start;
8564
	u64 i_size;
8565
	u64 cur_bytes;
Y
Yan Zheng 已提交
8566
	int ret = 0;
8567
	bool own_trans = true;
Y
Yan Zheng 已提交
8568

8569 8570
	if (trans)
		own_trans = false;
Y
Yan Zheng 已提交
8571
	while (num_bytes > 0) {
8572 8573 8574 8575 8576 8577
		if (own_trans) {
			trans = btrfs_start_transaction(root, 3);
			if (IS_ERR(trans)) {
				ret = PTR_ERR(trans);
				break;
			}
8578 8579
		}

8580 8581 8582
		cur_bytes = min(num_bytes, 256ULL * 1024 * 1024);
		cur_bytes = max(cur_bytes, min_size);
		ret = btrfs_reserve_extent(trans, root, cur_bytes,
8583
					   min_size, 0, *alloc_hint, &ins, 1);
8584
		if (ret) {
8585 8586
			if (own_trans)
				btrfs_end_transaction(trans, root);
8587
			break;
Y
Yan Zheng 已提交
8588
		}
8589

Y
Yan Zheng 已提交
8590 8591 8592
		ret = insert_reserved_file_extent(trans, inode,
						  cur_offset, ins.objectid,
						  ins.offset, ins.offset,
Y
Yan, Zheng 已提交
8593
						  ins.offset, 0, 0, 0,
Y
Yan Zheng 已提交
8594
						  BTRFS_FILE_EXTENT_PREALLOC);
8595 8596 8597 8598 8599 8600
		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			if (own_trans)
				btrfs_end_transaction(trans, root);
			break;
		}
C
Chris Mason 已提交
8601 8602
		btrfs_drop_extent_cache(inode, cur_offset,
					cur_offset + ins.offset -1, 0);
8603

J
Josef Bacik 已提交
8604 8605 8606 8607 8608 8609 8610 8611 8612 8613 8614 8615
		em = alloc_extent_map();
		if (!em) {
			set_bit(BTRFS_INODE_NEEDS_FULL_SYNC,
				&BTRFS_I(inode)->runtime_flags);
			goto next;
		}

		em->start = cur_offset;
		em->orig_start = cur_offset;
		em->len = ins.offset;
		em->block_start = ins.objectid;
		em->block_len = ins.offset;
8616
		em->orig_block_len = ins.offset;
J
Josef Bacik 已提交
8617
		em->ram_bytes = ins.offset;
J
Josef Bacik 已提交
8618 8619 8620 8621 8622 8623
		em->bdev = root->fs_info->fs_devices->latest_bdev;
		set_bit(EXTENT_FLAG_PREALLOC, &em->flags);
		em->generation = trans->transid;

		while (1) {
			write_lock(&em_tree->lock);
J
Josef Bacik 已提交
8624
			ret = add_extent_mapping(em_tree, em, 1);
J
Josef Bacik 已提交
8625 8626 8627 8628 8629 8630 8631 8632 8633
			write_unlock(&em_tree->lock);
			if (ret != -EEXIST)
				break;
			btrfs_drop_extent_cache(inode, cur_offset,
						cur_offset + ins.offset - 1,
						0);
		}
		free_extent_map(em);
next:
Y
Yan Zheng 已提交
8634 8635
		num_bytes -= ins.offset;
		cur_offset += ins.offset;
8636
		*alloc_hint = ins.objectid + ins.offset;
8637

8638
		inode_inc_iversion(inode);
Y
Yan Zheng 已提交
8639
		inode->i_ctime = CURRENT_TIME;
8640
		BTRFS_I(inode)->flags |= BTRFS_INODE_PREALLOC;
Y
Yan Zheng 已提交
8641
		if (!(mode & FALLOC_FL_KEEP_SIZE) &&
8642 8643
		    (actual_len > inode->i_size) &&
		    (cur_offset > inode->i_size)) {
8644
			if (cur_offset > actual_len)
8645
				i_size = actual_len;
8646
			else
8647 8648 8649
				i_size = cur_offset;
			i_size_write(inode, i_size);
			btrfs_ordered_update_i_size(inode, i_size, NULL);
8650 8651
		}

Y
Yan Zheng 已提交
8652
		ret = btrfs_update_inode(trans, root, inode);
8653 8654 8655 8656 8657 8658 8659

		if (ret) {
			btrfs_abort_transaction(trans, root, ret);
			if (own_trans)
				btrfs_end_transaction(trans, root);
			break;
		}
Y
Yan Zheng 已提交
8660

8661 8662
		if (own_trans)
			btrfs_end_transaction(trans, root);
8663
	}
Y
Yan Zheng 已提交
8664 8665 8666
	return ret;
}

8667 8668 8669 8670 8671 8672 8673 8674 8675 8676 8677 8678 8679 8680 8681 8682 8683 8684
int btrfs_prealloc_file_range(struct inode *inode, int mode,
			      u64 start, u64 num_bytes, u64 min_size,
			      loff_t actual_len, u64 *alloc_hint)
{
	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
					   min_size, actual_len, alloc_hint,
					   NULL);
}

int btrfs_prealloc_file_range_trans(struct inode *inode,
				    struct btrfs_trans_handle *trans, int mode,
				    u64 start, u64 num_bytes, u64 min_size,
				    loff_t actual_len, u64 *alloc_hint)
{
	return __btrfs_prealloc_file_range(inode, mode, start, num_bytes,
					   min_size, actual_len, alloc_hint, trans);
}

8685 8686 8687 8688 8689
static int btrfs_set_page_dirty(struct page *page)
{
	return __set_page_dirty_nobuffers(page);
}

8690
static int btrfs_permission(struct inode *inode, int mask)
Y
Yan 已提交
8691
{
L
Li Zefan 已提交
8692
	struct btrfs_root *root = BTRFS_I(inode)->root;
8693
	umode_t mode = inode->i_mode;
L
Li Zefan 已提交
8694

8695 8696 8697 8698 8699 8700 8701
	if (mask & MAY_WRITE &&
	    (S_ISREG(mode) || S_ISDIR(mode) || S_ISLNK(mode))) {
		if (btrfs_root_readonly(root))
			return -EROFS;
		if (BTRFS_I(inode)->flags & BTRFS_INODE_READONLY)
			return -EACCES;
	}
8702
	return generic_permission(inode, mask);
Y
Yan 已提交
8703
}
C
Chris Mason 已提交
8704

8705
static const struct inode_operations btrfs_dir_inode_operations = {
8706
	.getattr	= btrfs_getattr,
C
Chris Mason 已提交
8707 8708 8709 8710 8711 8712 8713 8714 8715
	.lookup		= btrfs_lookup,
	.create		= btrfs_create,
	.unlink		= btrfs_unlink,
	.link		= btrfs_link,
	.mkdir		= btrfs_mkdir,
	.rmdir		= btrfs_rmdir,
	.rename		= btrfs_rename,
	.symlink	= btrfs_symlink,
	.setattr	= btrfs_setattr,
J
Josef Bacik 已提交
8716
	.mknod		= btrfs_mknod,
8717 8718
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
8719
	.listxattr	= btrfs_listxattr,
8720
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
8721
	.permission	= btrfs_permission,
8722
	.get_acl	= btrfs_get_acl,
C
Chris Mason 已提交
8723
};
8724
static const struct inode_operations btrfs_dir_ro_inode_operations = {
C
Chris Mason 已提交
8725
	.lookup		= btrfs_lookup,
Y
Yan 已提交
8726
	.permission	= btrfs_permission,
8727
	.get_acl	= btrfs_get_acl,
C
Chris Mason 已提交
8728
};
8729

8730
static const struct file_operations btrfs_dir_file_operations = {
C
Chris Mason 已提交
8731 8732
	.llseek		= generic_file_llseek,
	.read		= generic_read_dir,
8733
	.readdir	= btrfs_real_readdir,
C
Christoph Hellwig 已提交
8734
	.unlocked_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
8735
#ifdef CONFIG_COMPAT
C
Christoph Hellwig 已提交
8736
	.compat_ioctl	= btrfs_ioctl,
C
Chris Mason 已提交
8737
#endif
S
Sage Weil 已提交
8738
	.release        = btrfs_release_file,
8739
	.fsync		= btrfs_sync_file,
C
Chris Mason 已提交
8740 8741
};

8742
static struct extent_io_ops btrfs_extent_io_ops = {
8743
	.fill_delalloc = run_delalloc_range,
8744
	.submit_bio_hook = btrfs_submit_bio_hook,
8745
	.merge_bio_hook = btrfs_merge_bio_hook,
8746
	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
8747
	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
8748
	.writepage_start_hook = btrfs_writepage_start_hook,
C
Chris Mason 已提交
8749 8750
	.set_bit_hook = btrfs_set_bit_hook,
	.clear_bit_hook = btrfs_clear_bit_hook,
J
Josef Bacik 已提交
8751 8752
	.merge_extent_hook = btrfs_merge_extent_hook,
	.split_extent_hook = btrfs_split_extent_hook,
8753 8754
};

8755 8756 8757 8758 8759 8760 8761 8762 8763 8764 8765 8766
/*
 * btrfs doesn't support the bmap operation because swapfiles
 * use bmap to make a mapping of extents in the file.  They assume
 * these extents won't change over the life of the file and they
 * use the bmap result to do IO directly to the drive.
 *
 * the btrfs bmap call would return logical addresses that aren't
 * suitable for IO and they also will change frequently as COW
 * operations happen.  So, swapfile + btrfs == corruption.
 *
 * For now we're avoiding this by dropping bmap.
 */
8767
static const struct address_space_operations btrfs_aops = {
C
Chris Mason 已提交
8768 8769
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
8770
	.writepages	= btrfs_writepages,
C
Chris Mason 已提交
8771
	.readpages	= btrfs_readpages,
8772
	.direct_IO	= btrfs_direct_IO,
8773 8774
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
8775
	.set_page_dirty	= btrfs_set_page_dirty,
8776
	.error_remove_page = generic_error_remove_page,
C
Chris Mason 已提交
8777 8778
};

8779
static const struct address_space_operations btrfs_symlink_aops = {
C
Chris Mason 已提交
8780 8781
	.readpage	= btrfs_readpage,
	.writepage	= btrfs_writepage,
C
Chris Mason 已提交
8782 8783
	.invalidatepage = btrfs_invalidatepage,
	.releasepage	= btrfs_releasepage,
C
Chris Mason 已提交
8784 8785
};

8786
static const struct inode_operations btrfs_file_inode_operations = {
C
Chris Mason 已提交
8787 8788
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
8789 8790
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
8791
	.listxattr      = btrfs_listxattr,
8792
	.removexattr	= btrfs_removexattr,
Y
Yan 已提交
8793
	.permission	= btrfs_permission,
Y
Yehuda Sadeh 已提交
8794
	.fiemap		= btrfs_fiemap,
8795
	.get_acl	= btrfs_get_acl,
8796
	.update_time	= btrfs_update_time,
C
Chris Mason 已提交
8797
};
8798
static const struct inode_operations btrfs_special_inode_operations = {
J
Josef Bacik 已提交
8799 8800
	.getattr	= btrfs_getattr,
	.setattr	= btrfs_setattr,
Y
Yan 已提交
8801
	.permission	= btrfs_permission,
8802 8803
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
J
Josef Bacik 已提交
8804
	.listxattr	= btrfs_listxattr,
8805
	.removexattr	= btrfs_removexattr,
8806
	.get_acl	= btrfs_get_acl,
8807
	.update_time	= btrfs_update_time,
J
Josef Bacik 已提交
8808
};
8809
static const struct inode_operations btrfs_symlink_inode_operations = {
C
Chris Mason 已提交
8810 8811 8812
	.readlink	= generic_readlink,
	.follow_link	= page_follow_link_light,
	.put_link	= page_put_link,
8813
	.getattr	= btrfs_getattr,
8814
	.setattr	= btrfs_setattr,
Y
Yan 已提交
8815
	.permission	= btrfs_permission,
J
Jim Owens 已提交
8816 8817 8818 8819
	.setxattr	= btrfs_setxattr,
	.getxattr	= btrfs_getxattr,
	.listxattr	= btrfs_listxattr,
	.removexattr	= btrfs_removexattr,
8820
	.get_acl	= btrfs_get_acl,
8821
	.update_time	= btrfs_update_time,
C
Chris Mason 已提交
8822
};
8823

8824
const struct dentry_operations btrfs_dentry_operations = {
8825
	.d_delete	= btrfs_dentry_delete,
8826
	.d_release	= btrfs_dentry_release,
8827
};